KR101271115B1 - System, Media, Program and Method on Generating Patent Risk Hedging Prediction Information - Google Patents

Abstract

The present invention relates to a method for generating patent risk hedging prediction information, a system for implementing the method, a program for performing the method, and a recording medium on which the program is recorded.
Implementation of the present invention enables the generation of systemically reliable and relevant patent risk hedging prediction information.

Description

Method for generating patent risk hedging prediction information, a system for implementing the method, a program for implementing the method, and a recording medium recording the program {System, Media, Program and Method on Generating Patent Risk Hedging Prediction Information}

The present invention relates to a method for generating patent risk hedging prediction information, a system for implementing the method, a program for executing the method, and a recording medium on which the program is recorded. A method for generating patent risk hedging prediction information that contributes to hedging of patent dispute risk using a dispute prediction model value, a system for implementing the method, a program for performing the method, and a recording medium on which the program is recorded.

In the era of the patent war, patent disputes are increasing worldwide. In the past, it was only at the level of constructing the dispute information in which the patent dispute occurred as a DB, analyzing the dispute information, transmitting the dispute occurrence information to a newsletter, or searching for the dispute information. These services were for patent disputes that occurred in the past and did not provide predictive information about future disputes. As a result, the disputed plaintiffs and the defendant companies are different from each other, and their products, technologies, and patent portfolios are different. Therefore, there is an urgent need to introduce a service that provides them with specialized dispute prediction information. Furthermore, there is an increasing need for patent information processing methods and systems for generating information capable of systematically hedging the risk of patent disputes, or for discovering cross-licensing targets or target patents.

The technical problem to be solved by the present invention is to present a method for generating patent risk hedging prediction information, a system for implementing the method, a program for implementing the method, and a recording medium on which the program is recorded.

In order to achieve the technical problem to be achieved by the present invention, (A) obtaining at least one disputed patent set and at least one non-disputed patent set including patents used in at least one or more kinds of patent disputes; (B) generate a dispute prediction element value for at least two dispute prediction elements preset for at least two dispute patents constituting the dispute patent set and at least two non-dispute patents constituting the non-dispute patent set; Making; And (C) the dispute patent and the non-dispute patent, wherein the dispute prediction element value is an explanatory variable value, and is different from the dispute patent grant value and the dispute patent grant value assigned to the dispute patent and are non-dispute. And establishing at least one dispute prediction model for generating at least one dispute prediction model by performing a predetermined statistical process using the non-dispute patent grant value assigned to the patent as a response variable value. A patent dispute prediction model generation method of a patent information system is presented.

The patents used in the patent disputes are used for patents used in patent disputes filed with the judiciary, patents used in patent disputes filed with the administration, patents used in warnings against patent infringers, and the exercise of patent rights against patent infringers. It is preferable that any one or more types of patents used and patents in which patent royalties have occurred occur.

The size of the non-disputed patent set is equal to or larger than the size of the disputed patent set, and the non-disputed patent set is extracted from the entire patent set while sharing any one or more of the statistical properties of the disputed patent set. It is preferably generated by any one method selected from the first non-dispute set generating method and the second non-dispute set generating method randomly extracted.

The dispute prediction element value is generated in each of the dispute patent set and the individual patent units constituting the non-dispute patent set for each dispute prediction element, and the dispute prediction element includes the self-property series, the other attribute series, and the field attribute for each individual patent. It is preferred to be associated with any one or more of the attribute sequences.

The dispute prediction element includes a citation related dispute prediction element, and when generating a dispute prediction element value for the citation related dispute prediction element, one or more citations of direct citation, indirect citation, latent citation, and chain citation are used. The method of using two or more of the direct citation, indirect citation, latent citation, and chain citation uses a first citation to independently process each citation type to generate a dispute prediction element value for a preset dispute prediction element. Using at least one of a second method of using a method and a second citation using a combination of two or more citations, and assigning a predetermined weight to each citation type to generate a dispute prediction element value for the preset dispute prediction element. Is preferably.

It is preferable that the dispute prediction element includes a dispute prediction element having at least a different period setting for the same kind of dispute prediction element, and the period setting is that at least one or more is set for the same type of milk powder. .

The dispute prediction element may be selected from at least one citation viewpoint dispute prediction element, at least one multi-dispute patent viewpoint dispute prediction element, at least one multi-dispute cause perspective dispute prediction element, and at least one multi-dispute technology group perspective dispute prediction element It is preferred to include any one or more.

The method for assigning the dispute patent grant value to the dispute patent may include a method for granting a first dispute patent value assigned differently according to the attributes of the dispute patent and a dispute patent grant value only if a dispute occurs regardless of the attributes of the dispute patent. It is preferable to use any one of the second dispute patent granting value granting methods for granting.

The attribute of the disputed patent is one or more of a plural dispute attribute, a joint defendant attribute, and a joint participation attribute. The plural dispute attribute is an attribute in which the dispute patent is related to at least two disputes. Is an attribute relating to a dispute in which the disputed patent raises a dispute against at least two or more defendants, and wherein the joint participation attribute is an attribute in which at least one or more other disputed patents are also related to the dispute related to the disputed patent. Do.

In the case of adopting the method for granting the first dispute patent granting value, the method for granting the dispute patent granting value is higher than the case where the dispute patent granting value has no plural dispute attribute when the dispute patent has plural dispute attributes. Granting the disputed patent if the disputed patent has the attributes of the joint defendant, or granting the disputed patent grant value higher than without the joint defendant. It is desirable to give the value lower than if it does not have a joint participation attribute.

In the case of adopting the method for granting the first dispute patent granting value, the method for granting the dispute patent granting value is the method for granting the dispute patent granting value when the dispute patent has the plurality of dispute attributes. The disputed patent, which is granted by referring to at least one of the total number of related disputes, the time-distributed attribute of all disputes, and the distribution-specific distribution attributes of all disputes, or when the dispute patent has the joint defendant attribute. The method of granting a grant value may be granted with reference to any one or more of the total number of defendants to which the dispute patent relates, the average defendant per dispute, and the statistical distribution attribute of the defendant per dispute. In the case of having a joint participation attribute, a method of assigning a dispute patent granting value may be performed in the dispute involving the dispute patent. Preferably, the patent is granted by referring to any one or more of the average share of the patent and the statistical distribution attribute of the average share.

Preferably, the statistical treatment is multiple regression analysis.

Preferably, the multiple regression analysis uses a nonlinear algorithm of machine learning series.

It is preferable that the nonlinear algorithm of the machine learning series uses an ensemble technique using a tree.

The dispute prediction model is generated by applying any one or more of the method for granting a first dispute patent grant value, the dispute prediction element group used, and the dispute patent set used differently according to the attributes of the dispute patent. desirable.

The dispute patent set used is preferably a dispute patent set that satisfies an attribute specified by the user.

(D) generating a dispute prediction model value for each patent for patents belonging to a patent set obtained using the dispute prediction model, wherein the obtained patent set is a whole patent set or a preset patent set. Or a patent set designated by the user or a related patent set related to the patent set designated by the user, wherein the generated dispute prediction model value is stored temporarily or permanently in response to the patent, or the dispute prediction It is desirable to send the model value to the requestor.

The generation of the dispute prediction model value is preferably generated according to a predetermined cycle or in accordance with a predetermined condition.

The patent information system which implements the method of any one of the above is proposed.

In order to achieve the technical problem to be achieved by the present invention, there is provided a recording medium on which a computer-readable program for implementing the method of any one of the above is recorded.

In order to achieve the technical problem to be achieved by the present invention, a computer-readable program for implementing the method of any one of the above is proposed.

In order to achieve the technical problem to be achieved by the present invention, at least two or more dispute prediction elements which are preset for at least two or more disputed patents constituting the disputed patent set and at least two or more non-disputed patents constituting the non-disputed patent set A dispute prediction element value generation unit configured to generate a dispute prediction element value for the; Regarding the dispute patent and the non-dispute patent, the dispute prediction element value is set as an explanatory variable value, and the dispute patent grant value and the dispute patent grant value assigned to the dispute patent are different from the dispute patent grant value. A dispute prediction model generation unit configured to establish at least one dispute prediction model generating a preset at least one dispute prediction model value by performing a predetermined statistical process using the non-dispute patent grant value as a response variable value; And a dispute prediction model value generation unit for each patent for generating a dispute prediction model value for each patent for patents belonging to a patent set obtained using the dispute prediction model. We present a patent information system that creates a dispute prediction model.

In order to achieve the technical problem to be achieved by the present invention, (a) (a1) (a11) first obtain a self-patent set including at least one patent, (a12) at least one or more having a predetermined relationship with the self-set Obtaining a target patent set or (a2) (a21) first obtaining a target patent set including at least one or more patents, and (a22) obtaining at least one self patent set having a predetermined association with the target set; (b) obtaining at least one dispute prediction model value for each patent for each patent constituting the target patent set; And (c) generating at least one or more dispute prediction information by using the dispute prediction model value for each patent.

The self patent set is a user management patent set which is a patent set created or managed by a user, or a system management patent set which is a patent set created or managed by the patent information system, and the self patent set uses a preset set limitation option. It is to be limited to, or to be divided into two or more partial patent set by applying a predetermined division criteria, the target patent set is a user management patent set which is a patent set created or managed by the user or the patent information system This is a system management patent set, which is a patent set that is created or managed, and the target patent set may be defined using a predetermined set limiting option, or may be divided into two or more partial patent sets by applying a predetermined division criterion. It is preferable.

The target patent set is a related patent relationship consisting of patents having a preset relationship with self patents constituting the self patent set, and the preset relationship is a citation-citation patent relationship, a similar patent group relationship in a text mining technique. In the patent classification, it is preferable that at least one of the similar technology patent relations be used.

Target patents constituting the target patent set may have a predetermined association with two or more self patents, the association may be quantified, and the quantification includes any one or more of an association frequency and an association strength. It is preferable that it is configured by.

The relevance frequency is the number of self patents having a predetermined relevance corresponding to the target patent, and the relevance strength is one or more of citation-cited citation relationship information, similar patent group relationship information, and similar technology patent relationship information. It is preferable that it is produced using.

Generating the relevance strength with the cited-cited citation relationship information preferably includes using one or more of cited-cited depth information and cited-cited type information.

Generating the relevance strength using the similar patent group relationship information is generated by using similarity information between the self patent and the target patent, and the similarity information is an extracted keyword extracted from the self patent and the target patent, Citation-cited information generated from reference information of the own patent and the target patent, at least one or more types of patent classifications extracted from the own patent and the target patent, and the patent on the patent classification system to which the patent classification belongs Preferably at least one step of the classification is generated including the utilization of any one or more of the upper patent classification.

Generating the relevance strength with the similar technology patent relationship information may correspond to at least one or more types of patent classifications included in the self patent and the target patent, the matching depth, the matching frequency, and the like in the patent classification system to which the patent classification belongs. It is preferably one that is generated utilizing any one or more of the matching sequences.

The dispute prediction model value for each patent may include at least one citation viewpoint dispute prediction element, at least one multi-dispute patent viewpoint dispute prediction element, at least one multi-dispute cause perspective dispute prediction element and at least one multi-dispute technology group perspective dispute prediction. It is preferable that it is generated using the dispute prediction element value of at least one dispute prediction element selected from the elements.

Obtaining the dispute prediction model value for each patent includes generating a dispute prediction model value for each patent in real time with respect to the target patent or in a patent dispute prediction model value DB for storing a previously generated dispute prediction model value for each patent. It is desirable to call.

The generated dispute prediction information may include dispute prediction information for each target patent corresponding to the acquired self-set, dispute prediction information for at least one target patent group corresponding to the acquired self-set, and target patent corresponding to the acquired self-set At least one of the dispute prediction information for the set, the dispute prediction information for each target patent corresponding to the at least one partial self-set, and the at least one dispute prediction information for the at least one target patent group corresponding to the at least one partial self-set is preferably one or more.

The dispute prediction information may be at least one of preset at least one dispute prediction information value, at least one dispute prediction analysis information, and at least one dispute prediction basis information.

The step (a12) is to obtain only a patent that meets a predetermined condition as the target patent, or the step (a22) is to obtain only a patent that meets a predetermined condition as its own patent. The set condition is preferably set by the user or set by the patent information system.

Preferably, the preset condition is any one or more of an owner condition, a relevance condition, and an individual patent attribution condition.

At least one of at least one target patent constituting the target patent set and at least one partial target patent set to which a predetermined division criterion or a predetermined selection criterion is applied to the target patent set has a predetermined ranking. The ranking is preferably generated according to a ranking generation rule using any one or more of the relevance frequency, the relevance strength, the dispute prediction model value, and user input relevance information.

Acquiring the target patent set in the step (a12) is to obtain the divided partial patents, and in this case, generating the dispute prediction information in the step (c) corresponds to the partial self-patent set. Is to be generated for each partial target patent set to be obtained or to obtain the self-patent set in the step (a22) is to obtain for each of the divided partial target patent set, in this case the dispute prediction of the step (c) Generating the information is preferably generated for each partial self patent set corresponding to the partial target patent set.

In the step (a), when the method of (a1) is selected, the acquisition of the self-patent set in the step (a11) is the acquisition of at least two self-patent sets, and the target patent of the step (a12) Acquisition of a set is obtained by each of two or more self-patent sets, and generating the dispute prediction information of step (c) is generated by a target patent set corresponding to the two or more self-patent sets. When the method of (a2) is selected, the acquisition of the target patent set in the step (a21) is the acquisition of at least two target patent sets, and the acquisition of the self patent set in the step (a22) is two or more of the It is obtained by the target patent set, and generating the dispute prediction information of the step (c) is generated for each self patent set corresponding to two or more target patent set It would be preferable.

The acquisition of the target patent set is at least one user patent set generated or selected by a user, and the association of the user patent set with the self patent set is related to at least one of the self patents constituting the self patent set and the user. It is preferable that at least one of the target patents constituting the patent set is capable of calculating any one or more of the citation-citation patent relationship, the similar patent group relationship in the text mining technique, and the similar technology patent relationship in the patent classification.

Generating the dispute prediction information of the step (c) is generated for each patent, or for each of at least two divided or selected self patent sets, or for each of at least two divided or selected target patent sets. The selection or division of the self-patent set or the selection or division of the target patent set may be selected or divided by a user or divided by predetermined selection criteria or division criteria of the system at all times. It is preferable that the above is applied.

The predetermined selection criterion or division criterion may be any one or more of an owner name, an attribute of an owner, a patent technology classification, and at least one predetermined evaluation value of a target patent constituting a target patent set.

The rights holder attribute may be any one or more of at least one attribute specified by the system and at least one attribute specified by the user, or the evaluation value may be any one or more of at least one qualitative evaluation element for the target patent. It is preferably a qualitative evaluation factor value evaluated using.

Obtaining a dispute prediction model value for each patent in the step (b) is performed on a secondary target patent constituting a secondary target patent set generated after performing a predetermined set operation on the obtained primary target patent set. The secondary target patent set is any one of deletion of the primary target patents constituting the primary target patent set, limitation on the primary target patent set, and addition of a new patent to the secondary target patent set. It is preferable that the above is performed and produced.

The generation of the dispute prediction information in the step (c) is to be generated by reflecting the user weight information received from the user, the user weight information for any one or more of the target patents constituting the target patent set Weight information for each target patent set by the user for each target patent, weight information for each attribute set for individual properties of the target patent, citation-to-cited weight information for the citation-to-cited relationship, and text mining relationship The text mining weight information and the similar technology patent relationship weight information set for the similar technology patent relationship in the patent classification are preferably one or more.

The dispute prediction model value is set to a dispute prediction element value as an explanatory variable value for dispute patents and non-dispute patents, and is provided differently from the dispute patent grant value and the dispute patent grant value assigned to the dispute patent, and the non-dispute patent It is preferable that the non-dispute patent grant value given to the response variable value is generated by performing a predetermined statistical process.

Preferably, the statistical processing uses a machine learning sequence algorithm using an ensemble technique using a tree.

In order to achieve the technical problem to be achieved by the present invention, there is proposed a patent information system for processing patent dispute prediction information, characterized in that the method of any one of the above.

In order to achieve the technical problem to be achieved by the present invention, there is provided a recording medium on which a computer-readable program for implementing the method of any one of the above is recorded.

In order to achieve the technical problem to be achieved by the present invention, a computer-readable program for implementing the method of any one of the above is proposed.

In order to achieve the technical problem to be achieved by the present invention, a self-patent set generation unit for obtaining a self-patent set including at least one patent; A target patent set generation unit generating or obtaining at least one target patent set having a predetermined relation with the self set; A dispute prediction model value obtaining unit which obtains at least one dispute prediction model value for each patent for the individual patents constituting the target patent set; And a dispute prediction information generation unit for generating at least one dispute prediction information by using the dispute prediction model value of each patent.

And a multiple relationship processing module that calculates a predetermined relationship between a self patent constituting the self patent set and a target patent constituting the target patent set. Is preferably at least one of a citation-cited patent relationship, a similar patent group relationship on a text mining technique, and a similar technology patent relationship on a patent classification.

In order to achieve the technical problem to be achieved by the present invention, (I) (I1) (I11) first obtain a self-patent set including at least one or more patents, (I12) at least one or more having a predetermined relationship with the self-set Obtaining a target patent set or (I2) (I21) first obtaining a target patent set including at least one or more patents, and (I22) obtaining at least one self patent set having a predetermined association with the target set; (J) obtaining at least one license prediction model value for each patent for each patent constituting the target patent set; And (K) generating at least one license prediction information by using the license prediction model value for each patent. The method provides a method for processing patent licensing prediction information of a patent information system.

In the step (I12), the target patent set is a post-application based on the best date of the self patent constituting the self patent set, or in the step (I22), the self patent set is a target constituting the target patent set. It is preferable that it is a prior application on the basis of the best date of a patent.

The self-patent and the target patent are related patent relationships consisting of patents having a predetermined relationship, and the predetermined relationship is a citation-to-citation patent relationship, a similar patent group relationship in a text mining technique, and a similar technology patent relationship in a patent classification. Any one or more of the target patents constituting the target patent set may have a predetermined relationship with two or more self patents, the relevance may be quantified, and the quantification is the relevance frequency and relevance intensity It is preferable that it is comprised including any one or more of these.

The relevance frequency is the number of self patents having a predetermined relevance corresponding to the target patent, and the relevance strength is one or more of citation-cited citation relationship information, similar patent group relationship information, and similar technology patent relationship information. It is preferable that it is produced using.

Generating the relevance strength with the citation-cited patent relationship information may include generating using at least one of citation-cited depth information and citation-cited citation type information or as the similar patent group relationship information. Generating the relevance strength is generated using similarity information between the self patent and the target patent, or generating the relevance strength with the similar technology patent relationship information is included in the self patent and the target patent. The at least one kind of patent classification, which is generated by utilizing any one or more of the matching depth, the matching frequency and the matching sequence on the patent classification system to which the patent classification belongs, the similarity information is the self-patent and the target Extracted keywords extracted from patents, the self-patent and the target feature Citation-cited information generated from fictitious reference information, at least one or more types of patent classifications extracted from the self-patent and the target patent, and at least one level higher patent classification of the patent classification on the patent classification system to which the patent classification belongs It is generated by using any one or more of the above, and generating the relevance strength with the similar technology patent relationship information is at least one kind of patent classification included in the self-patent and the target patent, the patent Preferably, the classification is generated by utilizing any one or more of the matching depth, the matching frequency, and the matching sequence on the patent classification system to which the classification belongs.

The generated licensing prediction information includes licensing prediction information per self patent corresponding to the target set, licensing prediction information for at least one self patent group corresponding to the target set, and licensing prediction information for the self patent set corresponding to the target set. The at least one sub-target set may be one or more of the licensing prediction information for each patent corresponding to the at least one sub-target set or at least one self-patent licensing prediction information corresponding to the at least one partial target set.

The licensing prediction information may be at least one of at least one preset license prediction information value, at least one license prediction analysis information, and at least one license prediction basis information.

Obtaining the target patent set in the step (I12) is to obtain only a patent that satisfies a predetermined condition as a target patent, or obtaining the target patent set in the step (I12) is the divided partial patent Acquisition for each set or acquisition of the target patent set in the step (I12) is for acquiring for each of two or more self-patent sets, and the preset condition is set by the user or the patent information system. The predetermined condition is any one or more of rights holder conditions, relevance conditions, owner sales volume, rights holder recent dispute suit information and individual patent attribute conditions, the acquisition of the target patent set is divided When obtaining the partial self patent set, generating the licensing prediction information of the step (d) When the partial target patent set corresponding to the partial self patent set is generated and the acquisition of the target patent set is obtained for each of the two or more self patent sets, the acquisition of the self patent set in the step (I) Is the acquisition of at least two self-patent sets, and the generating of the licensing prediction information of step (d) is preferably generated for each target patent set corresponding to the at least two self-patent sets.

Acquiring the self-patent set in the step (I22) is to obtain only a patent that satisfies a predetermined condition as a self-patent, or acquiring the self-patent set in the (I22) step is divided partial patents Acquisition of the self-patent set in the step (I22) is obtained by the target patent set of two or more, and the preset condition is set by the user or the patent information system. The predetermined condition is any one or more of rights holder conditions, relevance conditions, owner sales volume, rights holder dispute dispute information and individual patent attribute conditions, and the acquisition of the self-patent set When obtaining by the partial target patent set, generating the licensing prediction information of the step (d) is If the self-patent set corresponding to the partial target patent set corresponding to the partial target patent set is generated, and the acquisition of the self-patent set is obtained for each target patent set of two or more, the acquisition of the target patent set in the step (I) Is the acquisition of at least two or more target patent sets, and the generating of the licensing prediction information of step (d) is preferably generated for each of the own patent sets corresponding to the two or more target patent sets.

At least one of self patents constituting the self patent set and at least one partial self patent set to which at least one or more division criteria or at least one or more selection criteria are applied has a predetermined ranking, wherein the ranking is the relevance frequency, the It is preferably generated according to a ranking generation rule using any one or more of the relevance strength and the licensing prediction model value.

The generating of the licensing prediction information of the step (K) may be generated by at least two or more divided or selected target patent sets, or the generation of the licensing prediction information of the (K) step may be obtained from the user. It is generated by reflecting the user weight information, wherein the selection or division of the target patent set is selected or divided by the user or at least one of the methods divided by predetermined selection criteria or division criteria of the system at all times. The predetermined selection criterion or division criterion may be any one or more of an owner's name, an attribute of an owner, a patent technology classification, and at least one predetermined evaluation value of a target patent constituting a target patent set. User weight information is a target constituting the target patent set Weight information for each target patent set by the user for each or more patents of each patent, weight information for each attribute set for individual attributes of own patents, and citation-cited weight information set for a citation-cited relationship , At least one of text mining weight information set for the text mining relationship and similar technology patent relationship weight information set for the similar technology patent relationship in the patent classification.

Obtaining the dispute prediction model value for each patent in the step (J) is performed on the secondary target patents constituting the secondary target patent set generated after performing a predetermined set operation on the obtained primary target patent set. The secondary target patent set is any one of deletion of the primary target patents constituting the primary target patent set, limitation on the primary target patent set, and addition of a new patent to the secondary target patent set. It is preferable that the above is performed and produced.

The licensing prediction model value of each patent may include at least one citation perspective dispute prediction element, at least one multi-dispute patent perspective dispute prediction element, at least one multi-dispute cause perspective dispute prediction element, and at least one multi-conflict technology group perspective dispute prediction. It is preferable that it is generated using the dispute prediction element value of at least one dispute prediction element selected from the elements.

The dispute prediction model value may be assigned to a licensed prediction element value as an explanatory variable value for a licensed patent and a non-licensed patent, and is different from the licensed patent grant value and the licensed patent grant value granted to the licensed patent. It is preferable that the non-licensing patent grant value assigned to is generated by performing a predetermined statistical process using the response variable value.

Preferably, the statistical processing uses a machine learning sequence algorithm using an ensemble technique using a tree.

In order to achieve the technical problem to be achieved by the present invention, there is provided a patent information system for processing patent licensing prediction information, characterized in that the method of any one of the above.

In order to achieve the technical problem to be achieved by the present invention, there is provided a recording medium on which a computer-readable program for implementing the method of any one of the above is recorded.

In order to achieve the technical problem to be achieved by the present invention, a computer-readable program for implementing the method of any one of the above is proposed. In order to achieve the technical problem to be achieved by the present invention, the present invention includes at least one patent. A self patent set generation unit for obtaining or generating a patent set; A target patent set generation unit for obtaining or generating a target patent set including at least one patent; A dispute prediction model value obtaining unit which obtains at least one license prediction model value for each patent for each patent constituting the target patent set; And a licensing prediction information generation unit for generating at least one license prediction information by using the license prediction model value for each patent.

And a multiple relationship processing module that calculates a predetermined relationship between a self patent constituting the self patent set and a target patent constituting the target patent set. Is preferably at least one of a citation-cited patent relationship, a similar patent group relationship on a text mining technique, and a similar technology patent relationship on a patent classification.

In order to achieve the technical problem to be achieved by the present invention, (i) (i1) (i11) first obtain a target patent set including at least one or more patents, (i12) at least one or more of a predetermined relationship with the target set Obtaining a supplementary patent set or (i2) (i21) first obtaining a supplementary patent set including at least one or more patents, and (i22) obtaining at least one target patent set having a predetermined relationship with the supplementary set; (j) obtaining at least one dispute prediction model value for each patent for each patent constituting the complementary patent set; And (k) generating at least one risk hedging prediction information using the dispute prediction model value for each patent. The method provides a method for processing patent risk hedging prediction information of a patent information system.

In the step (i12), the complementary patent set is a prior application based on the best day of the target patents constituting the target patent set, or in the step (i22), the target patent set is the supplement that constitutes the target patent set. It is preferable that it is a post-application based on the best date of a patent.

The target patent set may be a patent set having a preset association with a self patent set including at least one patent, or a partial patent set of a patent set having a preset association with the self patent set, or more than the preset dispute prediction model value. Patents having a patent set consisting of patents, or a partial patent set of a patent set consisting of patents having more than the predetermined dispute prediction model value, or a patent having a predetermined dispute prediction information value or more in relation to the self-patent set A patent set consisting of patents or partial patent sets of patent sets consisting of patents having a preset dispute prediction information value in relation to the own patent set or a dispute prediction information value set in relation to the own patent set or more Patent set with It is preferable that it is a partial patent set of the patent set having more than a predetermined dispute prediction information value in relation to the own patent set.

The target patent and the supplementary patent are related patent relationships consisting of patents having a predetermined relationship, and the predetermined relationship is a citation-cited citation relationship, a similar patent group relationship in a text mining technique, and a similar technology patent relationship in a patent classification. Any one or more of the above, and the supplementary patents constituting the supplementary patent set may have a predetermined association with two or more target patents, the association may be quantified, and the quantification is the association frequency and the intensity of association. It is preferable that it is comprised including any one or more of these.

The relevance frequency is the number of target patents having a predetermined relevance corresponding to the supplementary patent, and the relevance strength is one or more of citation-cited patent relationship information, similar patent group relationship information, and similar technology patent relationship information. It is preferable that it is produced using.

Generating the relevance strength with the citation-cited patent relationship information may include generating using at least one of citation-cited depth information and citation-cited citation type information or as the similar patent group relationship information. Generating the relevance strength is generated using similarity information between the target patent and the supplementary patent, or generating the relevance strength with the similar technology patent relation information is included in the target patent and the supplementary patent. For at least one or more types of patent classifications, at least one of a matching depth, a matching frequency, and a matching sequence on the patent classification system to which the patent classification belongs, the similarity information is generated using the target patent and the supplement. Extracted keywords extracted from the patent, the target patent and the complementary feature Citation-cited information generated from the reference information, at least one or more types of patents extracted from the target and complementary patents, and at least one or more higher levels of the patent classifications on the patent classification system to which the patent classification belongs It is generated by using any one or more of the above, and generating the relevance strength with the similar technology patent relationship information for the at least one kind of patent classification included in the target patent and the supplementary patent, the patent Preferably, the classification is generated by utilizing any one or more of the matching depth, the matching frequency, and the matching sequence on the patent classification system to which the classification belongs.

The generated risk hedging prediction information includes risk hedging prediction information for each target patent corresponding to the complementary set, at least one risk hedging prediction information for each target patent group corresponding to the supplementary set, and a target patent set corresponding to the supplementary set. The risk hedging prediction information, the risk hedging prediction information for each target patent corresponding to the at least one partial complementary set, and the at least one risk hedging prediction information for the at least one target patent group corresponding to the at least one partial supplement set are preferably one or more.

The risk hedging prediction information may be at least one of at least one predetermined risk hedging prediction information value, at least one licensing prediction analysis information, and at least one licensing prediction basis information.

Acquiring the complementary patent set in the step (i12) is to obtain only a patent that satisfies a preset condition, or acquiring the complementary patent set in the (i12) step is divided partial patents Acquired by the set or the acquisition of the complementary patent set in the step (i12) is obtained by the target patent set of two or more, wherein the predetermined condition is set by the user or the patent information system The predetermined condition is any one or more of rights holder conditions, relevance conditions, owner sales size, rights holder dispute dispute information and individual patent attribute conditions, the acquisition of the complementary patent set is divided When obtaining by the partial target patent set, generating the risk hedging prediction information of step (d) Generated by the partial complementary patent set corresponding to the partial target patent set, and the acquisition of the supplementary patent set is obtained by each of the target patent sets of two or more of the target patent set of step (i). The acquisition may be acquisition of at least two or more target patent sets, and the generating of the risk hedging prediction information in step (d) may be generated for each complementary patent set corresponding to the two or more target patent sets.

In the acquisition of the supplementary patent set, the predetermined condition is that the system or user satisfies at least one attribute attribute of the supplementary patent selected or set by the supplementary patent, and the owner attribute information includes rights holder size information and rights holder. One or more of agency type information and whether the user is included in at least one group of related parties selected or designated by the user.

Acquiring the target patent set in the step (i22) is to obtain only a patent that satisfies a predetermined condition as a target patent, or acquisition of the target patent set in the (i22) step is divided partial complementary patent Acquisition of the target patent set in the step (i22) is obtained for each of two or more complementary patent sets, and the preset condition is set by the user or the patent information system. The predetermined condition is any one or more of rights holder conditions, relevance conditions, owner sales volume, rights holder recent dispute suit information and individual patent attribute conditions, the acquisition of the target patent set is divided When obtaining by the partial complementary patent set, generating the risk hedging prediction information of step (d) When the target patent set corresponding to the partial target patent set corresponding to the partial complementary patent set is obtained, and when the acquisition of the target patent set is obtained for each of the complementary patent set of two or more, the complementary patent set of the step (i) The acquisition may be acquisition of at least two complementary patent sets, and the generating of the risk hedging prediction information in step (d) may be generated for each target patent set corresponding to two or more complementary patent sets.

At least one of a target patent constituting the target patent set and at least one or more partial target patent sets to which at least one or more division criteria or at least one or more selection criteria are applied has a predetermined ranking, wherein the ranking is the relevance frequency, the It is preferably generated according to the ranking generation rule using any one or more of the relevance strength and the dispute prediction model value.

The generating of the risk hedging prediction information in the step (k) may be generated by at least two divided or selected complementary patent sets, or the generation of the risk hedging prediction information in the step (k) may be performed by the user. It is generated by reflecting the received user weight information, and the selection or segmentation of the supplementary patent set is either selected or divided by the user, or divided by the predetermined selection criteria or division criteria of the system at all times. The above is applied to, wherein the predetermined selection criteria or division criteria is any one or more of the name of the holder of the complementary patent, the attribute of the owner, patent technology classification, at least one predetermined evaluation value of the complementary patent set. The user weight information constitutes the supplementary patent set. Weight information for each complementary patent set by the user for each one or more of the supplementary patents, weight information for each attribute set for individual attributes of the target patent, and citation-cited weights set for the citation-cited relationship At least one of information, text mining weight information set for the text mining relationship, and similar technology patent relationship weight information set for the similar technology patent relationship in patent classification is preferable.

Obtaining the dispute prediction model value for each patent in the step (j) is performed on the secondary supplementary patents constituting the secondary supplementary patent set generated after performing a predetermined set operation on the obtained primary supplementary patent set. The secondary supplement patent set is any one of deletion of the primary supplement patent constituting the primary supplement patent set, limitation on the primary supplement patent set, and addition of a new patent to the secondary supplement patent set. It is preferable that the above is performed and produced.

The dispute prediction model value for each patent may include at least one citation viewpoint dispute prediction element, at least one multi-dispute patent viewpoint dispute prediction element, at least one multi-dispute cause perspective dispute prediction element and at least one multi-dispute technology group perspective dispute prediction. It is preferable that it is generated using the dispute prediction element value of at least one dispute prediction element selected from the elements.

The dispute prediction model value may be assigned to a licensed prediction element value as an explanatory variable value for a licensed patent and a non-licensed patent, and is different from the licensed patent grant value and the licensed patent grant value granted to the licensed patent. It is preferable that the non-licensing patent grant value assigned to is generated by performing a predetermined statistical process using the response variable value.

Preferably, the statistical processing uses a machine learning sequence algorithm using an ensemble technique using a tree.

In order to achieve the technical problem to be achieved by the present invention, there is provided a patent information system for processing patent risk hedging prediction information, characterized in that the method of any one of the above.

In order to achieve the technical problem to be achieved by the present invention, there is provided a recording medium on which a computer-readable program for implementing the method of any one of the above is recorded.

In order to achieve the technical problem to be achieved by the present invention, a computer-readable program for implementing the method of any one of the above is proposed.

In order to achieve the technical problem to be achieved by the present invention, a target patent set generation unit for obtaining or generating a target patent set including at least one or more patents; A supplementary patent set generating unit for obtaining or generating a supplementary patent set including at least one patent; A dispute prediction model value obtaining unit which obtains at least one dispute prediction model value for each patent for the individual patents constituting the supplementary patent set; And a risk hedging prediction information generation unit for generating at least one risk hedging prediction information using the dispute prediction model value for each patent.

And a multiple relationship processing module that calculates a predetermined relationship between a target patent constituting the target patent set and a complementary patent constituting the supplementary patent set. Is preferably at least one of a citation-cited patent relationship, a similar patent group relationship on a text mining technique, and a similar technology patent relationship on a patent classification.

The present invention has the following effects.

First, it is possible to create a system-reliable and reasonable patent dispute prediction model.

Secondly, it is possible to generate reliable and valid patent dispute prediction information systematically.

Third, it is possible to generate patent licensing prediction information that is reliable and feasible systemically.

Fourth, it is possible to generate systemic reliable and reasonable patent risk hedging information.

1 is an exemplary diagram of an environment for using a patent information system of the present invention.
FIG. 2 is a diagram showing a configuration of an embodiment of the patent information system of the present invention.
3 is a view of an exemplary configuration of a data unit of the patent information system of the present invention.
4 is a diagram of an exemplary configuration of a data processing unit of the present invention.
5 is a diagram illustrating an exemplary configuration of a patent information service support unit of the present invention.
6 is a diagram illustrating an exemplary configuration of a patent analysis information generation unit of the present invention.
7 is a diagram illustrating an exemplary configuration of an option selector of the present invention.
8 is a diagram illustrating an exemplary configuration of a key keyword DB of the present invention.
9 is a diagram illustrating an exemplary configuration of a system for generating patent dispute prediction information of the present invention.
10 is a diagram illustrating an exemplary configuration of a method for generating a dispute prediction element value for each dispute prediction element according to the present invention.
11 is a conceptual diagram for explaining an attribute of a cited patent set of the present invention.
12 is a conceptual diagram illustrating a forward cited patent set among the attributes of a cited patent set of the present invention.
13 is a conceptual diagram for explaining a back cited patent set among the attributes of a cited patent set of the present invention.
14 is a conceptual diagram illustrating a forward self citation patent set and a rear self citation patent set among attributes of a cited patent set according to the present invention.
15 is a conceptual diagram illustrating a cited generation obtained patent set among attributes of a cited patent set according to the present invention.
16 is a conceptual diagram illustrating a forward cited patent set and a back cited patent set as cited patent sets for the acquired patent set.
17 is a conceptual diagram illustrating a forward self citation patent set and a rear self citation patent set as cited patent sets for an acquired patent set.
18 is a conceptual diagram illustrating a citation generation obtained patent set as a cited patent set for an acquired patent set.
19 is a conceptual diagram illustrating a forward cited patent set and a back cited patent set as cited patent sets for a limited set of acquired patents under conditions. The forward cited patent set and the back cited patent set are also shown to be a subset of the forward cited patent set and the back cited patent set for the unrestricted acquired patent set as the limitation is placed on the acquired patent set. When conditions are placed on an acquired patent set, the forward cited patent set, the back cited patent set, the forward self cited patent set, the back self cited patent set, and the cited generation obtained patent set are all not limited to the forward cited patent set and the back cited set. A subset of the patent set, the forward self-citation patent set, the back self-citation patent set, and the citation generation obtained patent set.
FIG. 20 is a conceptual diagram illustrating a forward citation patent set and a limited backward citation patent set defined by applying conditions to the forward citation patent set and the backward citation patent set as the citation patent set to the obtained patent set. Like the limited forward citation patent set or the limited backward citation patent set, the forward self citation patent set, the back self citation patent set, and the citation generation obtained patent set are also limited, if the limitation is the forward self citation patent set, the rear self citation patent set. And a subset of the citation generation obtained patent set.
FIG. 21 is a conceptual diagram for explaining a first-type latent cited patent of the present invention. FIG.
FIG. 22 is a conceptual diagram for explaining a second kind of latent cited patent of the present invention. FIG.
Figure 23 is an exemplary conceptual diagram for explaining the first type chain citation patent of the present invention.
24 is a conceptual diagram illustrating an example of a chain citation patent of the present invention.
FIG. 25 is a diagram illustrating an embodiment of a method for generating a dispute prediction element value from a citation point of view of a patent dispute prediction information generation system according to the present invention. FIG.
FIG. 26 is a diagram illustrating an exemplary configuration of a method of generating a dispute prediction model of a system for generating patent dispute prediction information according to the present invention. FIG.
27 is a diagram illustrating a summary of the Gradient Boosting algorithm algorithm.
FIG. 28 is a diagram illustrating a summary of an algorithm proposed by Friedman (2002) to newly propose a stochastic gradient boosting algorithm.
29 is a diagram for explaining an exemplary concept of a dispute prediction model generation process of the system for generating patent dispute prediction information according to the present invention.
30 is an auxiliary diagram for explaining an overfit concept.
31 is a diagram illustrating an exemplary configuration of a method of generating a dispute prediction model through a boosting algorithm by a dispute prediction model generation engine of the patent dispute prediction information generating system according to the present invention.
32 is a diagram of an example of the stump generated by the boosting algorithm.
33 is a conceptual diagram illustrating a 5-fold cross validation technique.
FIG. 34 is a diagram of examples of stumps for categorical response variables. FIG.
35 is a diagram of an example of a stump for a regression model applied to a continuous response variable.
36 is a diagram illustrating an exemplary configuration of a method for generating and storing a dispute prediction model value by a system for generating patent dispute prediction information according to the present invention.
37 is a conceptual diagram illustrating that when two or more self patents have a relationship with two or more target patents, relationship information may correspond to each relationship.
FIG. 38 is a conceptual diagram illustrating that when one self-patent has a predetermined relationship with at least one target patent, each relationship may be assigned a predetermined weight.
FIG. 39 is a diagram illustrating an exemplary configuration of a method for generating dispute prediction information by the system for generating patent dispute prediction information according to the present invention. FIG.
40 is a diagram illustrating an exemplary configuration of a method in which a patent dispute prediction information generation system of the present invention generates weight information for each target patent and generates dispute prediction information in consideration of weight information.
FIG. 41 is a diagram illustrating an exemplary type of individual dispute prediction information generated by the patent dispute prediction information generating system of the present invention, and an exemplary sequence of generating dispute prediction information. FIG.
FIG. 42 is a diagram illustrating an embodiment of a method for generating dispute prediction information based on a target patent set corresponding to a divided self patent set by the system for generating patent dispute prediction information according to the present invention.
43 is a diagram illustrating an exemplary configuration of a method for generating dispute prediction information based on a divided target patent set by the system for generating patent dispute prediction information according to the present invention.
44 is a diagram illustrating an embodiment of a method for analyzing dispute prediction information in a system for generating patent dispute prediction information according to the present invention.
45 is a diagram illustrating an embodiment of a method of generating rating information according to a rating model of the patent dispute prediction information generating system of the present invention.
46 is a diagram illustrating an exemplary configuration of a method for generating new dispute propagation prediction information in a system for generating patent dispute prediction information according to the present invention.
FIG. 47 is a diagram illustrating an exemplary configuration of a dispute information alarm method of a patent dispute prediction information generating system according to the present invention. FIG.
48 is a diagram illustrating an exemplary configuration of a method of generating risk hedging prediction information of the patent dispute prediction information generating system of the present invention.
FIG. 49 is a diagram illustrating an embodiment of a method for generating cross-licensing prediction information of a patent dispute prediction information generating system of the present invention. FIG.
50 is a diagram illustrating an exemplary configuration of a system for generating patent licensing prediction information of the present invention.
FIG. 51 is a diagram illustrating an exemplary configuration of a method for generating licensing prediction information by a system for generating patent licensing prediction information according to the present invention. FIG.

EMBODIMENT OF THE INVENTION Hereinafter, the thought of this invention is demonstrated in detail, referring drawings.

As illustrated in FIG. 1, the patent information system 10000 of the present invention provides an information service to a user computer 100 through a wired / wireless network 200. The patent information system 10000 may be connected to at least one link system 300 and a wired / wireless network 200. An example of the link system 300 is a patent office or a raw data provider of each country that provides patent raw data. An example may be a system, a system of an enterprise information provider, or the like, and at least one or more systems that provide services by linking the services of the patent information system 10000 of the present invention may be examples.

As illustrated in FIG. 2, the patent information system 10000 includes a patent information service support unit including a data unit 1000, a data processing unit 2000, a search processing unit 3100, and a member management unit 3200. And a patent analysis information generation unit 4000, a patent dispute prediction system 5000, a patent licensing prediction system 6000, or a purpose-specific system unit 20000.

The data unit 1000 may include a patent data unit 1100, a non-patent data unit 1200, a key keyword DB 1300, a classification metadata DB 1400, a user data unit 1700, another DB 1800, One or more of the object-specific data unit 1900 and the like. The patent data unit 1100 includes a patent specification file unit 1110, a patent DB 1120, a patent classification DB 1130, other patent data DB 1140, and other classification DB unit.

The patent DB 1120 manages bibliographic matters of all patents, the specification body, and drawings for each field, and various fields (title, summary, prior art, claims, detailed description, etc.) constituting the specification body. ), Including key keywords extracted from Meanwhile, the patents may further include citation information as prior art documents for the patents. Taking the US patent data as an example, the cited information includes information included in the reference, including US patent document numbers, foreign patent document numbers, and indicators for non-patent documents. On the other hand, prior art research information by examiners or related persons of the Patent Office, and information on the examination when submitting the examiner's opinion are broad citation information. If there is forward citation information in a specific document, the specific document is a backward citation document from the viewpoint of the document included in the forward citation information. The document included in the forward citation information based on the specific document becomes the parent document and the specific document becomes the child document based on the parent document. It is obvious to those skilled in the art to process the information in the child-parent relationship with the DB, and a detailed description thereof will be omitted.

The bibliography of the patent document includes the issuing country information, various date information, various number information, at least one owner information, at least one inventor information, at least one patent classification information, at least one priority information, and the like. The date includes the filing date, the date of publication, the date of registration, and other dates. The various number information includes an application number, a public number, a registration number, an original number, and a priority claim number. The right holder information includes the applicant, the assignee, the patent owner, etc. If there is a change in the right holder and the change is managed, there may be information about the transferor and the assignee and information about the right holder. Priority information includes information such as priority claim number, date of claim, and country. On the other hand, when there is a split application, partial continuation application, continuation application, etc., information such as an original number, date of the original source is added. In addition, bibliographies may include representations, titles, abstracts, and index words. Meanwhile, the bibliographies processed include domestic family information (such as a split application, a change of application, or a part of a continuation application, a patent application that is related to a continuation application), overseas family information (an application that may be related to a treaty priority relationship, This can be. On the other hand, the text of the patent specification body may be further extracted by a keyword extracting method for each field or combination of fields constituting the main text through natural language processing or the like.

In addition to the common IPC, the patent classification information may have a unique patent classification for each country such as USPC, FT, FI, ECLA, etc. Thus, the patent classification has a hierarchical structure. Patent classification of the present invention, the patent classification includes IPC, USPC, FT, FI, ECLA. An index, also called a catchword, refers to a system in which at least one patent classification is associated with a word, phrase, or clause. Among the indexes, there is a catchword that processes the IPC, and there is an index to USPC issued by the US Patent and Trademark Office. The index also has a hierarchical structure like a patent classification. In the index, keywords corresponding to product names / part names / element descriptions are often embedded. It can be seen that the index makes it easy to find the patent classification. Information on catchwords for IPC provided by WIPO is available as ipcr_catchwordindex_20100101.zip at http://www.wipo.int/ipc/itos4ipc/ITSupport_and_download_area/20100101/MasterFiles/ as of March 2011. It can be seen that G06C 1/00 corresponds to ABACUSES in the file. In this case, ABACUSES is called an index corresponding to G06C 1/00. Accordingly, G06C 1/00 can be reverse mapped to Abacuses. On the other hand, as can be seen from the index ABARADING of the contents of the file, it can be seen that the catchword system has a hierarchical structure of at least one step. ABARADING shows that it is a two-tier hierarchy. The USPTO is an index for USPC and information about them is available at http://www.uspto.gov/web/patents/classification/uspcindex/indextouspc.htm as of March 2010. In the content of the uplink, it can be seen that the index Abrading is classified into three levels. It can be seen that the index Abrading corresponds to USPC 451/38, and there is a patent classification of a class other than Class 451 in the lower index of Abrading. Any matters related to the patent classification are stored in the patent classification DB 1130.

The core keyword DB 1300 stores information about the extracted core keyword. How to extract key keywords (including air pairs) from patent documents will be described later. The core keyword DB 1300 has a metadata key DB for each key keyword, and the meta information for each key keyword is stored therein. When a key keyword is extracted, a patent document number for which patent document is extracted for each keyword, a frequency in the extracted field, and a specific field of the specific patent document number are simultaneously calculated. The bibliographic information of the patent document number corresponds to various bibliographic information such as various date information, applicant information, inventor information, patent classification information, reference information, and the like. Therefore, a patent document number, bibliography, fields, and frequency may correspond to a specific key keyword, and this information becomes the basic content of the key keyword metadata DB 1310. On the other hand, if there is a basic content of the core keyword metadata DB 1310 as described above, the probability of mapping between increase and decrease rate data for each keyword, association analysis (basket analysis is an example) result data, and patent classification for each key keyword Data may be generated, and such data may also be contents of the core keyword metadata DB 1310. The mapping probability data between the patent classifications for each of the key keywords has a key keyword Ki, and the probability that Ki corresponds to at least one or more patent classifications Cj, P (Cj (Ki)), becomes the mapping probability. If Ki comes from patent documents P1, P2, P3 and P4, P1, P2, P3 is classified as C1, P4 is classified as C2, the probability that Ki corresponds to C1 is 3/4, and C2 The probability of matching is 1/4. That is, P (Cj (Ki)) = (the number of patent documents with Cj classification among the patent documents from which Ki was extracted)} / (the number of patent documents from which Ki was extracted). Meanwhile, when P (Cj (Ki)) is present, a set of Kis representing Cj may be extracted. In this case, P (Cj (Ki)) or Kis having a predetermined metric with respect to Cj of the KI may form this set.

The processed patent classification DB 1130 includes a classification metadata DB, in which meta information for each patent classification is stored. At least one kind and at least one patent classification correspond to a patent document, and the bibliographic information of the patent document number corresponds to various bibliographic information such as various date information, applicant information, inventor information, patent classification information, reference information, and the like. Will be. Therefore, a patent document number and a bibliographic matter may correspond to a specific patent classification, and this information becomes the basic content of the patent classification metadata DB. On the other hand, if there are basic contents of the patent classification metadata DB as described above, the increase / decrease rate data for each patent classification, the association analysis (an example of a shopping cart analysis), the result data, and the mapping probability data, share of the patent classification for each patent classification, Applicant data or inventor data having high patent index values may be generated for each patent index (such as a metric) such as an activity rate, and these data may also be contents of a patent classification metadata DB.

The user data unit 1700 stores all management information of a user who uses the patent information system 10000 and information generated or managed by the user. On the other hand, in the object-specific data unit 1900, the applicant DB 1910 including the applicant representative name information, etc., the rule data unit 1920 which stores various rules (patent indicators or analysis instruction syntax, etc.) information is stored; Information such as dictionaries for natural language processing (bandword dictionaries for translation, thesaurus dictionary, etc.) and other dictionaries (e.g., scientific and technical dictionaries, indexed dictionaries extracted from articles, etc.) 1930, the patents belonging to the standard patent pool and the like are stored in the standard patent pool data unit 1940. In addition, the purpose-specific data unit 1900 stores various types of data corresponding to the purpose.

As illustrated in FIG. 4, the data processing unit 2000 includes a key keyword generator 2100 for extracting at least one key keyword from a patent document, and classification metadata for generating various metadata for patent classification. Unit 2200, object-specific data generation unit 2300 for generating object-specific data, similar patent set generation unit 2400 for generating a patent set that is similar in content to a given patent document, and if there is an association analysis result, the association There may be a network data generation unit 2500 for generating one node and related information (edge information) between the associated nodes.

The patent information service support unit 3000 illustrated in FIG. 5 includes a search processor 3100 for performing a search process, a member manager 3200 related to member management, a service or data provided by the patent information system 10000, and the like. At least one or more user computers 100 by using a web service or SOA method (the user computer 100 of the present invention is not only a personal computer or a portable terminal such as a PC, but also a server or a system). 10000), which is the user computer 100 of the present invention.) A platform service provider for providing a platform service, an e-commerce service provider for providing an e-commerce service, and a community service provider for providing a community service. (8500), the billing processing unit (8600) for providing a billing service, to be able to comment in units of patent documents, the comment And a comment processing unit 8700 for processing any comment service, including performing a function of transmitting to the inventor or applicant of the patent document or providing the comment to the user computer 100 opening the patent document. do. The search processor 3100 may include one or more of a search engine 3110 and a DB query processor 3120. Since a typical DBMS supports an SQL search, the DBMS may perform a function of the DB query processor 3120.

6 is an exemplary diagram of a patent analysis information generation unit 4000 of the present invention. The patent analysis information generation unit 4000 obtains any one or more patent document sets or analysis target data sets obtained, and generates at least one preset analysis result for the obtained patent document set or analysis target data set. It performs the function. The acquisition of the patent document set or the analysis target data set for patent analysis is processed by the analysis target acquisition unit 4100. The patent analysis processor 4200 generates the analysis result. The patent analysis processor 4200 obtains an instruction for generating an analysis index value for at least one analysis index stored in the analysis index DB 4210, or an analysis command syntax stored in the analysis command syntax DB 4220. And obtain the result of analysis by applying to the patent document set or the analysis target data set obtained. Meanwhile, various options may be added in the patent analysis, which is performed by the analysis option processing unit 4300. The analysis option processing unit 4300 includes a data limiting option processing unit 4310 for defining a set of patent documents to be analyzed or data to be analyzed, and a display option processing unit for processing only what is displayed in the analysis results. 4320, and other option processing unit 4330 for processing various options for other patent analysis, and selection information for various options from the user computer 100, and the obtained option selection information is obtained from the patent analysis processing unit 4200. There is an option selector 4340 to transmit to. The analysis result transmitted to the user is generated by the analysis result reporting unit 4400. The reporting may be one or more of a table, chart, diagram (network diagram is one example), or a document in a predetermined format (eg, pdf, web page, etc.). Each of these generations is in charge of the table generator 4410, the chart generator 4420, the diagram generator 4430, and the report generator 4440. The set unit citation analysis 4500 applies at least one analysis index or analysis command syntax preset to at least two sets of patent documents to generate a patent analysis result. This will be described later.

7 shows a detailed configuration example of the option selector 4340 of the present invention. The option selector 4340 may select a country based on the country information among the owner's address and a time limiter 4340-1 which performs limitations related to the period of the patent document such as the application date, publication date, registration date, priority date, and best date. National limiting section 4340-2, applicant limiting section 4340-3 for processing limitations for at least one or more applicants, inventor limiting section 4340-4 for performing limitations for at least one or more inventors, at least A patent classification limiter 4431-5 which performs a limitation on one or more types of one or more patent classifications, by depth of patent classification, and by main patent classification and subpatent classification of a patent classification, and defines attributes of a specific field. A field attribute defining portion 4431-6 for performing a field, an individual field value defining portion for performing a specific limitation on a value of an individual field 4361-7, and an individual patent attribute value defining portion for performing a limitation on a specific attribute of an individual patent. (2111-8), Miscellaneous constraints 4434-9 that perform other constraints, such as citation depth, one or more of citation types (direct, indirect, latent, chain, etc.), and citation directions (front, back). . The field includes a field constituting a bibliographic matter, and may include an applicant, an inventor, a patent classification, and the like. For example, a company, an institution, a university, an individual among applicants, a limitation based on the size of a company such as a large company and a small and medium company, or a limitation based on an inventor address such as a domestic inventor or an overseas inventor among the inventors. . On the other hand, limitations on the applicants whose number of citations per patent are equal to or greater than a predetermined value, limitations on the inventors whose patent application growth rate is 25% or more in recent 5 years, and the like are limitations on individual field values. And based on the value of the property of the individual patent (such as the number of citations, patent grades, disputes, etc.), such as a patent with 10 or more citations, a patent within the top 10% of a patent rating evaluation, or a disputed patent. One limitation is an individual patent attribute value limitation.

8 is a diagram illustrating an exemplary configuration of the key keyword DB 1300 of the present invention. The key keyword DB 1300 includes a technology keyword DB 131, a product keyword DB 132, a syntax keyword DB 133, and a key keyword metadata DB 1310. The key keyword DB 1300 stores information on key keywords, air pairs, etc. extracted for each patent document, and the product keyword DB 132 includes various keywords or air pairs that characterize a product, and a phrase keyword DB 133. This includes technical problems (e.g., reducing weight, reducing the amount of hydrogen in metal to to action + action + material location of the object), and syntax for describing technical solutions. Information about a phrase keyword set representing a phrase having a specific attribute is stored in. Meanwhile, the core keyword metadata DB 1310 is not only described above, but also relations between keywords (as a result of association analysis). Relationship information can be generated.) Keyword relation information DB 1311 in which the information is stored, and keywords representing the patent classification for each patent classification. The representative keyword DB 1312 for each patent classification collected may be included.

9 is a diagram illustrating an exemplary configuration of a patent dispute prediction information generation system 5000 of the present invention.

The patent dispute prediction information generating system 5000 includes a dispute prediction engine 5100 for generating patent dispute prediction information, a dispute DB unit 5200 storing various data about disputes, a system and / or a user for patent disputes. The dispute prediction management unit 5400 for controlling the management information of the dispute prediction model engine for generating a dispute prediction model and a dispute prediction information analysis engine 5300 for analyzing the dispute prediction information may be included.

The dispute prediction engine 5100 generates a self patent set including a self patent comprising at least one patent generated or managed by the user or the patent information system 10000 or the patent dispute prediction information generation system 5000. Or a self-patent set generation unit 5110 to generate or specify a target patent set composed of at least one target patent having a predetermined relationship with a self-patent constituting the self-patent set, or generated or specified by the user. A target patent set generation unit 5120 for obtaining a target patent set composed of target patents, a dispute prediction model value for each patent generated by the dispute prediction model value generation unit 5530 for each patent, or a dispute prediction model generation unit And control to generate a dispute prediction model value of at least one or more designated patents through 5520 A dispute prediction model value acquisition unit 5130 that obtains a dispute prediction model value, a dispute prediction information generation unit 5140 that generates at least one or more preset dispute prediction information, and a dispute prediction information value that provides the generated dispute prediction information value The provision part 5150 is included.

The dispute prediction model generation engine 5500 may use the dispute prediction element value generation unit 5510, which generates a dispute prediction element value according to a dispute prediction element value generation rule set for each dispute prediction element, using at least one predetermined statistical algorithm. Dispute prediction model generation unit 5520 to generate a dispute prediction model described above, a dispute prediction model value generation unit 5530 for each patent to generate a dispute prediction model value for each patent using the dispute prediction model, the generated dispute prediction model value for each patent It may include a dispute prediction model value providing unit 5540 that provides.

Meanwhile, the dispute prediction element value generation unit 5510 may generate a dispute prediction element value generation unit 5510 for generating a dispute prediction element value for the quotation sequence dispute prediction element according to the generated attribute of the dispute prediction element value. The dispute prediction element values for the product / technology group dispute prediction element value generation unit 5510 and the owner, the applicant, the inventor, etc., which generate the dispute prediction element values for the product or technology family dispute prediction element, A subject-based dispute prediction element value generation unit 5510 to be generated, at least one or more trolls (patent troll) input or set by the user, patent technology classification, patent set, competitors or affiliates, standard patents, etc. There is a user input sequence dispute prediction element value generator 5510 that generates a dispute prediction element value for a user sequence dispute prediction element, such as a group.

The dispute DB portion 5200 includes a dispute occurrence patent DB 5210 for storing information on a dispute that has occurred, and the information stored in the dispute occurrence DB portion includes dispute unique identifier information, dispute processing court information, and at least one dispute. Target patent information (any patent information that can be specified by the patent number, such as patent number and / or bibliographic information corresponding to the patent number, and / or specification information, and / or drawing information, and / or progress information) At least one or more manuscript information, at least one or more defendant information, at least one or more litigation progress information, litigation result information, litigation depth information, and the like may be stored. Meanwhile, the dispute DB unit 5200 includes a dispute prediction element DB and a dispute prediction model DB for storing the dispute prediction element DB and rule data for generating the dispute prediction element value for each dispute prediction element (such as an SQL statement) and the dispute prediction model. And a dispute prediction element value DB 5220 for storing the dispute prediction element value for each patent, and a dispute prediction model value DB 5230 for storing the dispute prediction model value for each patent.

The dispute prediction management unit 5400 obtains patent dispute related data and manages a dispute data acquisition unit 5410 which processes parsing and the like, manages dispute related data input by a user, and a patent set input by other users. A user input patent group management unit 5432, dispute information generated by the dispute prediction engine 5100 in relation to a dispute, and the user using the patent dispute prediction information generation system 5000 in connection with a patent dispute prediction, It may include a dispute UI unit 5523 providing the generated dispute prediction information through a predetermined UI. The dispute prediction management unit 5400 is a dispute prediction user management unit managing a dispute prediction system management unit 5520 managing the patent dispute prediction information generation system 5000 and a user who uses the patent dispute prediction information generation system 5000. It may further include 5430. The dispute prediction system manager 5520 may further include a dispute prediction information batch generation unit 5251 which performs a batch process for generating preset dispute prediction information. The batching process may be performed by a predetermined owner unit, a predetermined patent classification unit, or a user-specific patent group managed by the dispute prediction management unit 5400, or a predetermined period or a predetermined condition achievement (update of dispute prediction model or dispute prediction element). Refers to a process for generating preset dispute prediction information according to value update, inflow of new dispute data, etc.). The dispute prediction user manager 5430 may further include a user input patent group manager 5432 for managing a patent group input by a user.

The dispute prediction information value providing unit 5150 includes a dispute information report generating unit 4440 that generates a report for providing patent dispute information or patent dispute prediction information in a document form, and generates the dispute information report. The report generated by the unit 4440 may be reported to the user through e-mail or in a predetermined manner through the dispute information input / output unit.

The dispute prediction information analysis engine 5300 may include a patent set dividing unit 5310 dividing a given patent set, an attack predicting patent group generating unit generating patent group information for which a patent attack is expected, and effectively hedging a patent dispute risk. It may include a risk hedging information generation unit 5330 for generating information, a cross-licensing information generation unit 5340 for finding a candidate patent group for cross licensing, or a licensing target information generation unit for generating licensing target information.

First, generation of the dispute prediction model will be described with reference to FIG. 10. The generation of the dispute prediction model is performed by the dispute prediction model generation engine 5500 of the present invention. The dispute prediction model generation engine 5500 obtains the dispute prediction element value generation rule information for each dispute prediction element (SL11), generates the input dispute prediction element value for each patent (SL12), and generates the dispute for each patent. The predictive element value is stored (SL13). In order to generate the dispute prediction model, derivation of a dispute prediction element must be preceded.

Derivation of dispute prediction factors can be derived through analysis of dispute patents. Patent disputed patents (disputed patents) are available for a fee or for free at online sites such as the patent practitioner at www.westlaw.com. Analyzing these patent dispute-related data, the disputed patents are distinguished from the non-dispute patents in the following aspects. A variable set called a dispute prediction element is generated based on the distinguished features.

First, disputed patents have a strong citation relationship. Patents in dispute tend to receive more citations than non-dispute patents, or to increase citations in recent years. In particular, there is often a strong citation relationship between the disputed patent and litigation defendant. In other words, the relationship between the defendant's patent and the disputed patent is established more than the relationship between the defendant's patent and the disputed patent. This citation relationship can be known by acquiring bibliographic matters such as information on the owner (applicant) of post-patent patents citing the disputed patent, and analyzing the bibliographic matters obtained. Reasons for citations to play an important role in relation to the subject of the dispute include: 1) patents tend to be reflected in the product, and the applicant for the post-applicant is required to protect the function of his / her product (especially to improve or improve). Many applicants apply for related patents, and 2) the patent holder of the post-application owner knows the patent of the pre-applicant owner and discloses it to IDS, etc. In many cases, 3) the pre-applicant also monitors the post-applicant's citation of his patent, and 4) the patent of the pre-applicant's patent, which receives a lot of citations from the post-applicant's patent, is reflected in the product of the post-applicant's Since there are many cases where there is a strong relationship (especially, the claims of the patentee of the first-applicant have broad scope of rights). If there is even more true) incorporated relationship has a dispute with extremely strong relation.

According to the relationship between the patent dispute and the citation as described above, the following candidate viewpoint for citation perspective dispute prediction factor can be derived. In view of the individual patents of the prior applicants, a citation viewpoint dispute prediction element candidate group as shown in Table 1 below may be considered as an example for a post-application patent citing an individual patent.

line Series code Citation perspective dispute prediction factor Remarks Total number A1 Total Blood Citations Last n years B1 Citations in the last n years (n can be more than one, the same applies in this table below) Self count C1 Self citations, self citations in the last n years Non-Self Count D1 Nonmagnetic citations, nonmagnetic citations in the last n years Average number E1 A1 to D1 per claim, the number of citations per citation, the number of citations per citation, and the number of citations per citation. Attribute value F1 A1, B1, D1, E1 based on the rights holder of the cited patent
A1, B1, D1, E1 based on the type of rights holder of the cited patent
A1, B1, D1, E1 based on the ratio of the right holder of the cited patent
A1, B1, D1, E1 based on patent classification of cited patent Increase or decrease rate G1 Increase or decrease rate of A1 ~ F1 for the whole period, increase or decrease rate of A1 ~ F1 for a certain period Reference H1 Number of references, number of patents in reference, number of papers compared to patent in reference, number of foreign patents in reference patent, number of foreign patents in reference / reference in foreign patents

Table 1 will be described. The right holders are classified into corporations, institutions, universities, and individuals, which in turn can be divided into domestic and foreign countries, and corporations can be divided into large and small businesses, and corporations and individuals are private entities, and institutions and universities are public entities. It can be divided roughly, and it can be divided into a multi-application subject and a small-application subject. The Maintenance Fee information issued by the US Patent and Trademark Office contains information on whether a right entity is a large entity, and in other countries, a multi-application / multiple rights entity and a small-application / sub-rights entity based on the number of patent applications or patent registrations of the right holder / applicant. Can be divided into The division of corporations, institutions, and universities can be grasped by organization type information included in the name of the applicant / authority (name = organization name + organization type). For example, name Samsung Electronics Co. Ltd., Samsung Elelctronics becomes the organization name and Co. Ltd. becomes an organization type. The ratio of the right holder type may be a company owner / total owner, (corporate + individual) / (institution + university), etc. Self refers to the case where the owner of an individual patent and the owner of a cited patent are the same (so-called Self Citation), while Non-Self means only considering cases where the owner of a particular patent differs from the owner of the cited patent.

On the other hand, such abbreviated notation as "A1, B1, D1, E1" and A1-D1 introduced because there are too many notation items are demonstrated. In terms of A1, B1, D1, and E1 based on the rights holder of the citation patent, A1 is the total number of right holders of the citation patent, B1 is the total number of right holders of the citation patent for the last n years, and D1 is the right holder of the citation patent. Among them, the total number of rights holders of individual patents and other rights holders, E1, refers to a value obtained by dividing A1, B1, and D1 values by the number of claims of an individual patent. Thus, the shorthand notation consists of "element part + range of reference part". In the above example, the element portion becomes the "owner of the cited patent," and the scope portion is the range of the reference portion that will constitute a candidate for dispute prediction element based on the assignee of the cited patent.

On the other hand, the classification of patents is based on the types of patent classifications (IPC, USPC, FT, FI, ECLA), the level of the patent classification hierarchy (for example, subclass or main group in the case of IPC), main patent classification and The dispute prediction element candidate value may be generated for each one or a combination of two or more of the scope of use of the subpatent classification. For example, if a patent contains both USPC and IPC, only the main patent classification is used, and USPC uses only the main patent classification, and IPC is the class prediction unit candidate for a total of three categories, such as in the subclass unit and the main group unit. You will be able to generate a value. The same applies to the processing of patent classification as far as dispute prediction element candidates and dispute prediction elements are concerned.

There are four types of citations in the present invention. The first is direct citation, which corresponds to a citation of depth one. When patent P2 is included in the reference (quotation) information of patent P1, P2 becomes a parent of P1, and P1 becomes a child of P2. In this case, P2 is a direct citation of P1 and P1 is a direct citation of P2. The second is indirect citation, which corresponds to citation depth 2 or higher. If there is P3 in the reference of P2, P3 is an indirect citation of the citation depth 2 of P1. Third is latent citation. Fourth is chain citation. All kinds of citations are divided into forward and backward. The dispute prediction element values or the dispute prediction element candidate values for the various dispute prediction elements to the dispute prediction element candidates of the citation viewpoints as shown in Table 1 are classified by the four types of citations (by subtypes according to the four types of citations (for example, 1 kind of late citation, 2 kind of late citation, 1 kind of chain citation, 2 kind chain citation, etc.) It may be generated by a simple summing or adding with different weights set according to the four types of citations. Of course, it may be reasonable that the dispute prediction factor value or the dispute prediction factor candidate value is generated by separating the front and the rear.

In addition, self-citation of consultation refers to the case in which a patent is cited by an applicant who has a pre-established close relationship (eg, affiliated relationship, etc.) with the company or the company. Self-citation is a case in which the self and the reference have a predetermined common denominator, and the common denominator may include the applicant common, the inventor common, and the common patent classification.

The cited patent set generation unit 5121 may be configured to directly generate a cited patent set including at least one direct cited patent, an indirect cited patent set including at least one indirect cited patent, and at least one potential cited patent for the own patent SSi. Generating at least one of a potential cited patent set comprising, and a chain cited patent set comprising at least one chain cited patent.

Reference information related to the direct citation of the present invention is 1) given by the applicant or owner of Pi (applicant cited), 2) given by the examiner (cited by the examiner), 3) given through prior art research (prior art research) Citations), 4) any one or more of the citations cited in the examiner's Pi review (consideration citations). In the citation patent set generation unit 5121, the citation patent set generation is generated by considering any one or more types of indirect citation information, latent citation information, and chain citation information.

Subsequently, the self-patent set, cited patent set, forward cited patent set, back cited patent set, forward cited patent set, back self cited patent set, back cited patent set, and cited generation obtained patent sets are obtained with reference to FIGS. 11 to 20. Explain.

In FIG. 11, the I1 to I6 patents constitute a set of obtained patents. I1 cites P1, I2 cites P2, I3 cites P3 and P4, I4 cites P4 and I4 belonging to the patent set, I 4 cites I 5, I5 cites I6 Quoting. On the other hand, I1 is cited by C1, and I2 is cited by C2 to C4. At this time, with respect to the cited patent sets I1 to I6, P1 to P5 and I5 and I6 become forward cited patent sets, and C1 to C4 and I4 and I5 become back cited patent sets. As can be seen from Fig. 12, when the cited patent set is limited to I1 to I4, the forward cited patent set is P1 to P5 and I5. On the other hand, as can be seen in Figure 13, when the cited patent set is limited to I1 to I3, the back cited patent set is C1 to P4.

14 illustrates the concept of a self citation diagram. I4 to I6 belong to the set of obtained patents, I5 is forward citation at I4, I6 is forward citation at depth 2, I6 is forward citation, I4 is backward citation, I5 is I5 Is the back citation and I4 is the back citation of depth 2. Therefore, when the obtained patent set is limited to I1 to I4, I5 becomes the forward self citation patent set of depth 1, and I6 becomes the forward self citation patent set of depth 2. In the forward self-cited patent set, if I1 to I6 are the same rights holder, the relationship between I5 and / or I6 and I4 is likely to be a similar technology since I5 and / or I6 constituting the forward self-cited patent set are patents of the same right owner. It is high, and I4 is likely to be a technology that improves or improves I5 and / or I6, so that it is possible to know the trend of technology development (continuity, concentration) of a specific owner. This is also the case for the back self-cited patent set. That is, for the patents included in the cited patent set, the patents related to the forward self cited patent set and the back self cited patent set become strongly related, and in the owner's patent portfolio, continuity, improvement, extensibility and comprehensiveness. Representation of the last name, etc. This is the case in which the cited patent set is viewed from the owner's point of view, and the cited patent set is defined by the inventor unit or the cited patent set by the patent classification (represents the technical field). That is, the patents of the same inventor among the patents of a specific inventor in the cited patent set, the patent generation of the citation generation patent set and the back self-cited patent set represent the continuity of technology development or research, based on the inventors. can do. Viewing these results is a unique effect of the collective unit citation analysis of the present invention.

Next, a citation generation acquisition patent set is demonstrated. The cited generation obtained patent set is composed of a patent group that receives at least one citation of the patents constituting the obtained patent set. Of the patents constituting the cited patent set, I1, I2, I4, and I5 constitute the cited generation obtained patent set. On the other hand, if the restriction to exclude the patent constituting the acquisition patent set of the patents constituting the citation generation acquisition patent set, which is one of the special limitations among the citation generation acquisition patent set, the limited citation generation acquisition patent set as shown in FIG. Will be created. If the right holders of I1 to I6 are the same, and the applicants of C1 to C4 are different from the right holders of I1 to I6, I2 becomes the most cited patent by the other right holders, and becomes an important patent among the cited generation obtained patent sets. There is a high possibility, and the inventor of such an important patent is also likely to become an important inventor.

Subsequently, the self-patent set, cited patent set, forward cited patent set, back cited patent set, forward self cited patent set, back self cited patent set, and citation generation obtained patent set are obtained with reference to FIGS. 16 to 20. Explain. In FIG. 16, an input set (IS) means a patent set obtained, a parent of input set (PIS) refers to a forward cited patent set, and a child of input set (CIS) refers to a back cited patent set. In FIG. 17, a cross-over parent of input set (CPIS) refers to a front self-cited patent set, and a cross-over child of input set (CCIS) refers to a rear self-cited patent set. Patents belonging to Cross-over Parent of Input Set (CPIS) and Cross-over Child of Input Set (CCIS) also belong to IS. 18 refers to a citation occurrence acquisition patent set as COIS (Citation Occurred IS). The feature of the backward cited patent set for all patents belonging to the IS and the back cited patent set for all patents belonging to the COIS is the same.

19-20 illustrate the case where limitations occur. This limitation is equally applicable to the cited patent sets generated through indirect citations, latent citations or chain citations, as well as direct citations. When a specific restriction is placed on the acquired patent set IS, a SIS (Specified Input Set) is generated. A PSIS (Parent of Specified Input Set), a limited forward cited patent set for the SIS, and a CSIS (Child of Specified Input Set), a limited back cited patent set, are generated. 20 is a type of a limited forward citation patent set generated when the forward citation patent set and the backward citation patent set are generated for the acquired patent set IS, when preset limitations are performed on the forward citation patent set and the backward citation patent set. SPIS (Specified Parent of Input Set) is created first, and then the specification is applied to the forward cited patent set and SCIS (Specified Child of Input Set), a kind of backward cited patent set. Will be.

Limitation may be limited through the option selector 4340 of the present invention, and FIG. 7 illustrates an embodiment of the configuration of the option selector 4340. The above limitations include time limitations, country limitations, applicant limitations, inventor limitations, patent classification limitations, field attribute limitations, individual field value limitations, individual patent attribute value limitations, and other limitations. 4340-1), country limiter 4340-2, applicant limiter 4340-3, inventor limiter 4340-4, patent classification limiter 4431-5, field attribute limiter 4431-6 ), The individual field value limiting unit 4341-7, the individual patent attribute value limiting unit 2111-8, and the other limiting unit 4341-9. The time limit can be selected from to to, and there may be various criteria for the period, such as an application date, a publication date, and a registration date. A country's limitation may be the country listed in the priority claim (the nationality of the rights holder) or the nationality of the inventor (in the inventor's address information), and the status represents various current status of the patent, including disclosure, registration, invalidation, and extinction. . In the case of patent classification, the selection of the patent classification type for IPC, USPC, FT, FI, ECLA, the selection of whether it is the main patent classification or the sub patent classification, and the depth (in the case of IPC, section, class, subclass, main) Group, 1 dot subgroup, n dot subgroup, etc.) On the other hand, the selection of numerical values includes the number of co-inventors, co-applicants, citations, patent classifications, patent grades, patent scores There may be a choice of numerical values for each field to be measured, calculated or obtained for individual patents such as.

The method of performing the limitation may include a period limitation, a country limitation, a rights holder limitation, an inventor limitation, a patent classification limitation, and the like, which is a time limitation portion 4340-1, a country limitation portion 4340-2, and an applicant limitation portion. 4340-3, inventor limiter 4340-4, and patent classification limiter 4431-5. When limiting the applicant or the like, a list of rights holders included in the acquired patent set is shown, and a user using the patent information system 10000 can select at least one rights holder. The same holds true for the inventor limitation and the patent classification limitation.

Next, special limitations are introduced. The first is field attribute qualification. For example, when a specific owner is classified as a patent monster or a competitor, a patent monster or a competitor is a special field attribute. This is particularly important because there is a high risk of patent disputes or licensing requirements, especially when quoting a large number of patents of right holders classified as patent monsters. The right holder is the same field as the applicant, and the property field of the applicant is not only a patent monster or a competitor (the patent monster or competitor may be different for a specific user and may be designated by a user). There may be a variety of attributes, such as research institutes, firms), owner size (large and small firms), owner attributes (multiple firms, many key patent holders, and companies with high patent quality indices). It will be apparent that the inventor and the like can be given various attributes as well as the owner. This field attribute limitation is performed by the field attribute limiter 4431-6 of the present invention. On the other hand, when the field attribute is present as a value, such limitation is performed by the individual field value limiter 4431-7 in the present invention. On the other hand, for a specific patent set, the field conditions are limited by field values of individual fields such as two or more co-applicants, three or more co-inventors, two or more patent classifications, five or more citations, and so on. Only the patent group that satisfies may be extracted. This restriction is performed by the individual patent attribute value limiter 4431-8. On the other hand, when there are other limitations such limitations are performed by the other limiting portion (4341-9) the present invention. It is a function of the option selector 4340 to perform the above limitations and the like comprehensively. The above limitation may be performed for both the acquired patent set and the cited patent set, and when the generation of the obtained self-patent set obtains the first obtained set and processes the obtained first obtained set as the acquired patent set, Also for the primary acquisition set, when the generation of the cited patent set generates the primary target patent set, the primary target patent set may be limited. The case where this limitation is performed is well illustrated in FIGS. 19 and 20. In FIG. 19, in the case of further limiting to the acquired patent set, the forward cited patent set is also limited than the forward cited patent set for the unrestricted acquired patent set for the limited obtained patent set, and the back cited patent set for the limited acquired patent set. It is shown that it is more limited than the back cited patent set for the unrestricted acquired patent set. Meanwhile, FIG. 20 illustrates that when the forward cited patent set and the back cited patent set are generated with respect to the obtained patent set, the forward cited patent set and the back cited patent set may be limited to generate a limited forward cited patent set and the back cited patent set. Show that there is. Special limitations include keyword limitations. That is, when core keywords are extracted for individual patents, a patent set may be limited to keyword-related conditions such as including or excluding at least one specific keyword.

Subsequently, the potential citation will be described in more detail with reference to FIGS. 21 to 22. The latent cited patent includes a first latent cited patent and a second latent cited patent.

The first type of potential cited patent is a patent owned by self patent (SSi) and a common parent patent (for example, P (SSi), a direct cited patent of SSi), without having a direct citation relationship with self patent Refers to a patent in which the cited patent is disclosed and available on a predetermined reference date, or the self-patent is disclosed and available on an established reference date of the cited patent. First class latent citations are patents such as LCP1 and LCP4 in FIG. 21. In order to find the first class latent cited patent (LCP1) among the first type latent cited patents, the patent set directly cited by the self-patent (direct citation with citation depth 1 as well as indirect citation patents with citation depth 2) is the target. Patents cited therein may be found earlier than their filing date. In order to find the late class 1 latent cited patent (LCP4) among the first type of late cited patents, the patent set directly cited by the self-patent (direct citation with citation depth 1 as well as indirect cited patents with citation depth 2) is also included. See patents cited after the publication date of their patents.

The second type of potential cited patents are self-patented patents that do not have a direct citation relationship with their own patents among patents having self patents (SSi) and common parent patents (for example, P (SSi), direct cited patents of SSi). The patent does not have a possibility of obtaining because the cited patent is not disclosed on a predetermined reference date, or refers to a patent that could not be obtained because its own patent was not disclosed on the established reference date of the cited patent. The second type of potential cited patents are patents such as LCP2 and LCP3 in FIG. 22. In order to find a forward second class latent cited patent (LCP2) among the second latent citation patents, the patent set directly cited by the self-patent (direct citation with citation depth 1 as well as indirect citation patents with citation depth 2) is the target. For patents cited), the filing date is earlier than the filing date of the own patent, and the publication date is later than the filing date of the own patent. In order to find the late class 2 latent cited patent (LCP3) among the second type of late cited patents, the patent set directly cited by the self-patent (direct citation with citation depth 1 as well as indirect cited patents with citation depth 2) is the target. For patents cited), the application date is later than the filing date of the patent and before the publication date of the patent.

It is preferable that the predetermined reference date be treated as an application date. Strictly speaking, the reference date is a date for inputting a reference for each patent reference (e.g., when the examiner inputs a reference, the input date when the examiner inputs a reference, or when the prior art research document is input, the date of the prior research document investigation, etc.). On the other hand, in the United States, where there is an IDS scheme, the submission date may vary for each reference). On the other hand, if the filing date is the reference date, the international filing date becomes the usual filing date when entering the domestic phase through the PCT international application, but the filing date is more likely to be treated as the filing date at the domestic stage in the case of international filing. Would be more reasonable.

Next, the chain citation patent will be described in more detail with reference to FIGS. 23 to 24. The chain citation patent includes a first kind chain citation patent and a second kind chain citation patent.

The first type chain cited patents may be subject to direct cited patents (direct citation with citation depth 1 as well as indirect cited patents or late cited patents with cited depth 2) which are patents directly cited by SSi. Among the patents cited by the direct-cited patent and the self-patent patent may have a common parent patent (for example, WCP1), or a direct-cited patent which is a patent citing self patent (SSi) directly. And a patent (eg, WCP4) in which the patent cited by the direct-cited patent and the self-patent may have a common parent patent. The first chain citation patent is well illustrated in FIG.

The second type chain cited patents include patents which are directly cited by SSi (for example, WCP2) (direct citation having a citation depth of 1 as well as indirect citing patents or potential citing patents having a citation depth of 2). At least one citation depth (e.g., WCP3) and the self-patent having a common parent patent (e.g., P (SSi)). Refers to a directly cited patent (WCP3) of at least one citation depth of a patent (WCP2) directly cited by SSi and a patent (WCP2) directly cited by the self patent (SSi). Meanwhile, the second type chain citation patent includes a direct citation patent (eg, WCP6) and at least one citation depth of at least one citation depth of a patent (eg, WCP5) that directly cites the self patent (SSi). In the case of having a common parent patent (for example, P (SSi)), at least a patent (WCP5) that directly cites the magnetic patent (SSi) and a patent (WCP5) that cites the magnetic patent (SSi) directly. Refers to a direct cited patent (WCP6) of one or more cited depths.

In order to find a chain citation patent using only the direct citation patent, a chain citation patent set is composed of the patents citationd directly by the own patent and the self patent, and then a patent having a citation-citation relationship is extracted from the chain citation patent set. Among the extracted patents, the filing date is earlier than the patent, and the filing date is later than the patent. On the other hand, among the extracted patents, when the self-patent is cited or when the self-patent becomes a citation, the first patent is a chain citation patent, and when there is no direct citation relationship with the own patent, the second patent is obtained. In order to find the chain citation patent using only the indirect citation patent, a chain citation patent set may be composed of self patents and indirect citation patents, and the same information processing may be performed. On the other hand, in order to find the chain citation patent with only the potential citation patent, a chain citation patent set may be composed of self patents and indirect citation patents, and the same information processing may be performed. On the other hand, the self-patent may comprise the chain citation patent set of any two or more of direct citation patents, indirect citation patents, and latent citation patents, and then perform the same information processing.

In all of the patents related to the above citations, it would be reasonable to set the front and rear on the basis of the filing date between the self-patent and the cited patents, but the best date (in the case of a priority claim application, the priority date is the best, and there is no priority claim) In this case, the normal filing date may be the best date.

The dispute prediction element (candidate) value for each dispute prediction element (candidate) as shown in Table 1 related to the citation in the present invention may be any one or more of the direct citation, the indirect citation, the latent citation, and the chain citation. You can create by type. Furthermore, each dispute prediction element may subdivide the dispute prediction element by the depth of indirect citation and / or by the type of potential citation or chain citation, and may generate the dispute prediction element value by the segmented dispute prediction element. . On the other hand, if not subdivided, the range used in generating the actual dispute prediction model among the depth of the indirect citation and the type of the potential citation may be selected by the user or the administrator of the patent dispute prediction information generation system 5000.

When there is a specific patent set (Self Set, SS), the patent set formed of the patent group cited by the patent included in the patent set may be a parent patent set (Parent Set), the patent satisfying a specific condition ( For example, patents citing patents included in the patent set within the last three years or patents including a specific applicant or patent classification, etc.) may also be a patent set (Parent Set). Any combination of patents in which a registered patent is included in this set of prior patents becomes a group of prior patents. A patent set formed of a patent group citing a patent included in the patent set or a patent group citing a patent included in the patent set while satisfying specific conditions may be a post-application patent set (Child Set). Any combination of patents included in the post-application patent set is a post-application patent group. The cited viewpoint dispute prediction element value may be generated for each specific patent set, for each prior patent set, or for each subsequent patent set. If the specific patent set B is a post-patent patent set of the patent set A, then A becomes a Self Set, B becomes a Child Set, and if the specific patent set B is a pre-patent patent set of a patent set C, then C Becomes Self Set, and B becomes Parent Set. On the other hand, if the specific patent set B is a post-patent patent set of the patent set A, the parent set of B does not necessarily become A due to the nature of the citation-to-cited relationship. It is common for a subset of B's Parent Set to be A.

FIG. 25 illustrates one embodiment method for generating a dispute prediction element value for each dispute prediction element in a citation view with respect to Table 1. FIG. The dispute element value generation engine acquires a patent for which a dispute prediction element value is generated (SL21), obtains a directly cited patent of the acquired patent, generates a dispute prediction element value related to the direct citation, or indirectly cites the obtained patent. Obtain a patent and generate a dispute prediction factor value of an indirect citation, obtain a potential citation patent of a received patent, generate a dispute prediction factor value of a late citation, obtain a chain citation patent of the obtained patent, A dispute prediction element value is generated (SL22), and the generated citation related dispute prediction element value is stored (SL23).

Second, there is a multi-dispute-induced patent that participates in a number of patent disputes in a set of disputed patents, such as many disputed patents causing more than one dispute. Rather than causing even disputes over all registered patent rights, there are only a few disputed patents involved in a dispute, and in particular, several multi-dispute-induced patents cause a number of disputes. The multi-dispute causing patent can also be known by analyzing patent dispute information. Obtain a patent for which a patent dispute has occurred by patent dispute, obtain a patent number related to a patent dispute, and more than the predetermined number of participations in a dispute among the patent numbers becomes a multi-dispute-inducing patent. Such dispute-producing patents become first-class multi-dispute-inducing patents.

On the other hand, there are patents that have not been disputed but are highly likely to be disputed. For example, at least one or more patent groups of subsidiaries, affiliates, affiliates of patent monsters or patent monsters, or patent holders controlled by patent monsters, patents contained in at least one standard patent pool, or patents held by multi-dispute creators Examples include patents in the group or newly incorporated into them (whether or not they can be found by processing current assignee information). The patents belonging to this example are referred to as a second type multi-dispute inducing patent.

Meanwhile, patents such as a reissue patent, a multi-family amount patent, a multi-family amount increase patent, a multi-citation patent, and a patent with a high increase rate of citation are referred to as a third type multi-dispute-induced patent.

Multi-dispute-induced patent perspective disputes based on individual patents Whether the first, second and / or third-class multi-dispute-induced individual patents are identified on a per-prediction basis. Predictive element values may be generated. Per-patent multi-dispute-prone patent perspective Dispute prediction factor values may be treated as 1 if applicable, such as whether a Reissue patent is applicable, or 0 if not applicable. It may also be a value other than 1 (for example, the amount of the family is an integer, the rate of increase may be a rational value). Meanwhile, the first, second and / or third type multi-dispute-prone patent viewpoint dispute prediction element values may be generated for each of the multi-dispute-prone patent viewpoint dispute prediction elements in units of patent groups instead of individual patents. That is, when there is a multi-dispute-induced patent group consisting of n patents, the multi-dispute-induced patent viewpoint dispute prediction element value may be generated for each of the n-patent-induced patent viewpoint dispute prediction elements. In this case, the generated multi-dispute-prone patent viewpoint dispute prediction element value may be any one of a count (including summation) value, an average value (arithmetic mean, geometric mean, etc.) and a predetermined function value. For example, when there is a patent group consisting of 100 patents, when the number of Reissue patents is 10, 10 may be a count value. On the other hand, even if ten of the same Reissue patent is smaller the population (n), the population includes a relatively large number of Reissue patents. In other words, the density of Reissue increases, and the higher density increases the possibility that the patent group is related to the dispute. On the other hand, the density may be measured as an average value.

Furthermore, the citation point of view dispute prediction element for each patent of the first, second and / or third kind of multi-dispute-induced patents or for each of the first, second, and / or third kind of multi-dispute-induced individual patent groups. Of course, it is possible to generate candidate values.

In view of the multi-dispute-induced patents, dispute prediction element candidates as shown in Table 2 may be exemplarily presented.

line Series code Multi-dispute-Prone Patent Perspective Dispute Prediction Factor Candidates Remarks Total number A2 Re-disputes (participation-1), families, change in owners, or non-inventor assignee changes Last n years B2 A2 for the last n years (e.g., number of re-disputes in the last n years, number of families in the last n years, change in the rights holders in the last n years, and change in assignee nonees in the last n years) (n may be one or more, The same applies) Increase or decrease rate C2 The amount of increase and decrease of standard A2-B2 for the whole period, increase and decrease rate of standard A2-B2 for specific period Whether D2 Reissue patent status, foreign patent status, troll patent status Bibliography E2 No. of Application Claims, No. of Registration Claims, No. of Drawings, No. of Pages, No. of Patents, No. of Heterogeneous Patents Based on Specific Patent Classification Levels, No. of Co-Applications, No. of Co-Inventors, No. of Registrations, No. of Retention Periods, No. of Foreign Inventors, No. of Foreign Inventors Percentage, priority claims, and retention period Standard or Membership Pool E2 Inclusion of standard patent pool, inclusion of standard patent pool, inclusion of affiliated patent pool Multiple disputes F2 Number of dispute participation, number of joint participation disputes, number of joint participation patents per dispute, total defendants, defendants per participation dispute Multiple disputes G2 F2 in the last n years, F2 increase or decrease in the previous period, F2 increase or decrease in a specific period Dispute distribution H2 First dispute year, last dispute year, least disputed year Progress information I2 Judgment by Judgment, Number of Judgments by Judgment, Request for Priority Examination, Difference between filing date and request, Difference between filing date and registration date

Third, there is a multi-dispute inducer. The multi-dispute cause information is obtained by applying various conditions such as a predetermined frequency or a predetermined increase rate in accordance with the patent dispute information and the owner or plaintiff of the patent number included in the patent dispute information. One or more polydisputers can be extracted. The multi-dispute inducer that analyzes and generates the conflict information is called the first multi-dispute inducer. On the other hand, the right holder who owns a patent monster or a predetermined number of patents in at least one or more patent pools, or holds a patent that satisfies a predetermined condition (recent number, high growth rate), etc., is called a second type of multi-dispute inducer. In addition, among the right holders, the proportion of Reissue patents, multi-family patents, multi-family patents, multi-patent patents, and increased citations among the patent holders determined by the owner's entire patent portfolio or at specific times or under certain conditions. The right holder who owns a patent group with a high proportion of high-speed patents is called a third type multi-dispute right holder.

From the viewpoint of the multi-dispute inducer, a dispute prediction factor candidate may be presented as an example.

line Series code Multi-Dispute Provoker Perspective Conflict Predictor Candidate Remarks Multidispute Relevant A3 Number of dispute participation patent holders, total number of dispute participation patent holders, ratio of dispute participation patents among patent holders Multidispute Experience B3 Number of disputes filed by right holder, number of disputes filed against number of disputes filed by right holder Relevance of Standard Patent Pool Participants C3 No. of Participants' Standard Patent Pool Participation Patents, No. of Patents Participating in Standard Patent Pool Participants, and Ratio of Participant's Patents in Standard Patent Pool Participants Last n years D3 A3-C3 in last n years Increase or decrease rate E3 A3 ~ C3 increase / decrease in recent n years, A3 ~ C3 increase / decrease in recent n years Registered Troll Relevance F3 Whether the owner is a registered troll Owner Authority Attributes G3 Whether it is a large company, an overseas company, or whether the owner is a company / institution / university / individual Owner Patent Portfolio Properties H3 Total number of claimed claims, number of claims per patent application, number of claims per inventor, total number of patents registered, total number of claimed claims, patent registration rate, total number of effective patents, valid patent remaining rate, average duration of valid patents, patents Party Citation Index, Technology Influence Index, Technology Index, Technology Progress Measurement Index, Innovation Speed Index (isi), Scientific Linkage, Science Index, Scientific Linkage Index, International Knowledge Flow (IKF), Technology Dependency, Technology Independence, Co-Application Rate , Co-applicants

When there is a Set consisting of n patents, one or more of the count value, average value, and a predetermined function value of a patent in which the multi-dispute initiator is the owner of the n patents is set as the multi-dispute viewpoint perspective dispute prediction element value. can do. For example, when there is a patent group consisting of 100 patents, if the number of patents in which the multi-dispute-provoker of AA is the owner is 10, 10 may be a count value. On the other hand, even if there are ten patents in which the same multi-dispute inducer is the owner, the smaller the population (n), the larger the number of patents in which the multi-dispute inducer is the owner. In other words, the density of multi-dispute triggerers becomes higher, and the higher density increases the likelihood that the patent group will be related to the dispute. On the other hand, the density may be measured as an average value.

Furthermore, a patent group or preset conditions (specific period, specific patent classification, or specific search term or other condition) generated by the entire patent of the multi-dispute inducer of the first, second and / or third multi-dispute inducers For each patent group that satisfies the above, the cited viewpoint dispute prediction element candidate value may be generated.

Fourth, there is a multi-dispute technology group in which litigation occurs frequently. In addition, there exists a technology group that causes a lot of disputes based on a patent portfolio owned by the rights holder and a technology group that does not. The technology portfolio of the rights holder may be different, and subjective importance of the rights holder may be different. The right holder may normally have two or more patents for two or more technology groups (for example, a patent classification), and the patents included in each technology group of the same rights holder may have different disputes and the number of disputes filed. .

For example, in the US patent dispute information, it can be seen that among the patent classifications of dispute-inducing patents, a number of patent disputes occur in IPC standards G06F and A61K compared to other patent classifications. The dispute-inducing patent includes at least one or more types of at least one patent classification, and the patent classification is processed to extract at least one multi-dispute technology group. The multidispute technology group may be a patent classification included in the IPC, USPC or other patent bibliography. Meanwhile, the multi-dispute technology group information for various hierarchies can be generated in the hierarchical structure of the patent classification system. For example, when there is a patent classification of H04B 7/26, the parents of H04B 7/26 (2 dot Subgroup) are in sequence H04B 7/24 (1dot Subgroup), H04B 7/00 (Main Group), H04B (Subclass), H04 (Class), and H (Section). Therefore, at least one of the most frequent patent classifications may be extracted based on the patent classification of the disputed patent group for each layer such as n dot subgroup, main group, subclass, and class. At this time, whether to extract the most frequent patent classification using only the main patent classification or to extract the most frequent patent classification including the sub classification may vary according to the system setting or the user's selection. Of course, the depth of the patent classification hierarchy (which may be different for each type of patent classification system such as n dot subgroup in Section in the case of IPC and n dot Subgroup in Class in the case of USPC) is also determined by the system setting or user selection. It may vary.

From the perspective of the multi-dispute technology group, a dispute prediction element candidate as shown in Table 4 may be presented as an example.

line Series code Multi-dispute Technology Group Perspective Conflict Prediction Factor Candidate Remarks Multidispute Technology Group Relevance A4 Number of disputes over patent classification criteria,
Number of patents filed
Patent classifier disputes,
Patent classifier dispute defendants,
Ratio of number of dispute occurrences to total number of patents based on patent classification,
Ratio of disputed patents to total patents based on patent classification
Ratio of disputes to total patents,
Ratio of disputed defendants to total number of patents based on patent classification Relevant patent group limited multi-dispute technology group relevance B4 Number of patent disputes based on patent group-specific patent classification
Number of patents for disputes
Number of patent dispute holders
Number of disputed defendants
The ratio of the number of times a dispute arises out of the total number of patents
Patent Group Based on Patent Group Limited Patent Classification Criteria Ratio of Disputed Patents
Patent holder-limited patent classification criteria The ratio of the number of disputed applicants to the total number of patents,
Patent group-specific patent classification criteria The ratio of disputed defendants to total patents Owner Technology Sector Portfolio Attributes C4 Owner patent classification share,
Owner patent classification criteria concentration rate,
Owner patent classification criteria activity rate,
Owner Patent Classification Criteria H3
Patent Holder Criteria Average Duration of Registration per Patent Last n years A4 to C4 in the last n years Increase or decrease rate A4 ~ C4 increase or decrease based on the last n years, A4 ~ C4 increase or decrease based on the last n years Technology attribute The economic life of the technology

In Table 4, the patent classification is classified by patent classification type (eg, IPC, USPC, etc.), patent classification depth (eg, IPC, class, subclass, main group, n dot subgroup, etc.), main patent classification, or all patents. It may be any one of the classification or a combination of two or more. On the other hand, in Table 4, the patent group based on the patent group limited patent group refers to a dispute prediction element generated only for the patent classification included in the patent group held by the owner of the specific patent Pi. For example, if there is an A owner who holds m patents in the C1 patent classification and n patents in the C2 patent classification, the dispute prediction element values corresponding to B4 are processed separately by m and n. Of the m patents of the right holder A, two are the disputed patents, and if there are one disputed patent among the n patents, the number of disputed patents based on the patent group limited patent classification criteria of the right holders owned by the right holder of the rights holder A The factor value is 2 for C1 and 1 for C2, and the dispute prediction factor value of the number of disputed patents against the total number of patents based on the patent group-limited patent classification is 2 / m for C1 for C1. , C2 is 1 / n.

On the other hand, the dispute prediction element value generation unit 5510 is at least one patent consisting of at least one patent or for every patent belonging to the patent DB 1120 in advance or for a predetermined period or based on the achievement of a predetermined condition Dispute prediction element values of one or more or all dispute prediction elements of Tables 1 to 2 may be generated for each set. On the other hand, the dispute prediction element value generation unit 5510 is in real time dispute prediction of any one or more or all of the dispute prediction elements of Table 1 to Table 2 by at least one or more patent sets or individual patents composed of at least one or more patents You will be able to create element values. The dispute prediction element value generation unit 5510 may include at least one or more patents configured by all patent owners belonging to the patent DB 1120 or at least one or more patents in advance or at predetermined periods or based on achievement of preset conditions. For each patent owner included in a patent set, dispute prediction element values of one or more or all dispute prediction elements in Table 3 may be generated. On the other hand, the dispute prediction element value generation unit 5510 is any one or more or all of the disputes in Table 3 with respect to patent holders of each patent holder or individual patents included in at least one or more patent sets consisting of at least one patent in real time Dispute prediction element values of the prediction elements may be generated.

In addition, the dispute prediction element value generation unit 5510 may include at least one of patents of all patents belonging to the patent DB 1120 or at least one or more patents in advance or at predetermined periods, or based on achievement of preset conditions. Generate dispute prediction element values for any one or more of Table 4 or for all dispute prediction elements in each of the patent classifications of patents included in one or more patent sets, or by patent classification based on the owners of patents included in at least one patent set. You can do it. On the other hand, the dispute prediction element value generation unit 5510 predicts any one or more or all disputes in Table 4 for patent classification of each patent classification or individual patents included in at least one patent set composed of at least one patent in real time. Dispute prediction factor values for the components may be generated. The object of the information processing of the dispute prediction element value generation unit 5510 may be generated for at least one or more patent classification depths on the patent classification system to which the patent classification belongs, with respect to a preset patent classification type. The processing may be performed by patent classification type, and the information processing may be performed using only the main patent classification or using the subpatent classification together. For example, in the case of US patents, if IPC and USPC are indicated, IPC processes to generate dispute prediction element values in subclass and main group units, and USPC generates dispute prediction element values in class units. You could do it. In this case, only the main patent classification is used in the case of IPC, and the sub-patent classification is also considered in the case of USPC (for example, if there is a C1 main patent classification and a C2 sub patent classification in the Pi patent, the dispute is predicted based on C2). When generating an element value, the dispute prediction element value may be generated by processing Pi as an object of information processing).

On the other hand, when the dispute prediction element value generation unit 5510 processes information about a specific patent classification on a patent classification basis, patents corresponding to lower patent classifications of the specific patent classification are included in the information processing of the specific patent classification. Can be processed. For example, when H04B 7/24 is a multi-dispute technology group, the dispute prediction element value generation unit 5510 may be a parent of a patent group including H04B 7/24 as a main IPC (of course, H04B 7/24 is a parent). If there is a child patent classification as a patent classification, it includes a patent group consisting of patents covering all of the child patent classifications, which would be natural due to the nature of the hierarchy of the patent classification. For each patent group that meets a condition (a specific period, a specific patent classification, or a specific search word or other condition), a dispute prediction element value for each preset dispute prediction element is generated.

The relationship between the conflict prediction element and the candidate for the conflict prediction element will be described. Dispute predictor candidates are essentially conflict predictor elements. When a dispute prediction model is generated for a dispute prediction element candidate, a dispute prediction element candidate that does not contribute to the dispute prediction model or falls below a predetermined level may be discarded, and the remaining ones may be a dispute prediction element. However, conflict prediction elements that contribute very little (typically, very rarely, have zero contributions to the model for each conflict prediction element in the generation of the model) or are very minimal. Even if included in the dispute prediction model, such a dispute prediction element has a small contribution (importance), so even if such a dispute prediction element is included, the dispute prediction model value is hardly changed. Thus, all dispute prediction element candidates may be dispute prediction elements. If a conflict prediction element with a low contribution (importance) is included, a dispute prediction element value should be generated for the dispute prediction element, reflected in the model generation, and a dispute prediction model value is generated. Since the element value must be reflected, it may be uneconomical and inefficient in terms of computing power consumption. However, if the computing power is sufficient, it may be a desirable method to use many dispute prediction factors.

Dictionary-based prior statistical analysis may be performed on the dispute prediction elements shown in Tables 1 to 4. The dispute prediction model generation engine 5500 is configured with respect to any one or more of the dispute prediction elements, predetermined statistics for the dispute prediction element value for individual patents belonging to each of the patent sets for each dispute patent set and non-dispute patent set; Generate statistical processing information. The statistics are numerical values for the mean, variance, standard deviation, distribution attribute, etc. of the dispute prediction element value, and the statistical processing information may include visualized information such as a comparison graph for statistical analysis. For certain dispute prediction elements, if the value of the dispute prediction element for the disputed patent set and the value of the dispute prediction element for the non-dispute patent set are the same or extremely similar, these dispute prediction elements may be excluded from the beginning in the generation of the dispute prediction model. (Dropped out at the candidate stage)

Next, a concrete method of generating a dispute prediction model using the dispute prediction element value generated by the dispute prediction element value generation unit 5510 will be described. FIG. 26 illustrates one embodiment method of generating a dispute prediction model. The dispute prediction model generation engine 5500 acquires (SL31) at least one or more dispute patent sets and at least one non-dispute patent set including patents used in at least one or more types of patent disputes, and constitutes at least one dispute patent set. Generate a dispute prediction element value for at least two dispute prediction elements preset for at least two dispute patents and at least two non-dispute patents constituting a non-dispute patent set (SL32); For example, the dispute prediction element value is set as an explanatory variable value, and the dispute patent grant value assigned to the dispute patent and the dispute patent grant value are different from each other, and the non-dispute patent grant value assigned to the non-dispute patent is set as a response variable value. Statistical processing is performed to generate at least one preset dispute prediction model (SL33). Hereinafter, powdering is performed.

In order to generate a dispute prediction model, a dispute patent set and a non-dispute patent set are generated first. A dispute patent set is a set of disputed patents. The non-dispute patent set refers to patents that do not cause a dispute, and since the majority of non-dispute patents are used, the non-disputed patent set used for generating a dispute prediction model is separately configured through sampling. Sampling for generation of a non-dispute patent set for all non-dispute patents determines the sample size and then uses 1) random extraction, 2) stratified extraction, 3) or a predetermined statistical extraction method. The larger the sample size of the non-dispute patent set is, the better, but the decision is made in consideration of the size and computing power of the dispute patent set, but it is necessary to set the same or larger than the size of the dispute patent set. When the stratification extraction method is selected, proportional stratification extraction is appropriate. The proportional stratification extraction is any one of time distribution information such as the registration year or filing year of the disputed patent constituting the disputed patent group, and the technology distribution information based on the patent classification. It may be implemented in consideration of one or more. On the other hand, if the patent owner has caused a large number of disputes during the proportional stratification extraction, the stratification extraction in which the number of disputed patents by the right holders of the disputed patents constituting the disputed patent group is considered, such as extracting a large number of non-disputed patents of the rights holder, is partially. Some stratified extractions may be implemented, taking into account some of the number of disputed patents by year of dispute.

The dispute prediction model generation engine 5500 divides the disputed patent set and the non-disputed patent set, respectively, or combines them, and divides them into at least two or more, so that the divided first dispute patent set and the divided first non-disputed patent set are dispute prediction. The model may be used for model generation, and the divided second dispute patent set and the divided second non-dispute patent set may be used for verifying the generated dispute prediction model. At this time, the size of the first dispute patent set and the first non-dispute patent set is preferably larger than the second dispute patent set and the second non-dispute patent set.

The dispute prediction element value generation unit 5510 generates or obtains a dispute prediction element value for each dispute prediction element for the dispute patent and the non-dispute patent constituting the dispute patent set and the non-dispute patent set.

The dispute prediction model generation unit 5520 allocates a dispute patent response variable value to a patent belonging to the dispute patent set, and assigns a dispute patent response variable value different from the dispute patent response variable value to a patent belonging to the non-dispute patent set. . The method for allocating the dispute patent response variable value is largely divided into 1) a method of matching the same value to all dispute patents, and 2) a method of matching different values according to the attributes of the dispute patent. In the former case, the statistical model uses a classification model, and the latter case uses a regression model. In the latter case, the classification model may be used depending on the design of the response variable.

According to the nature of the disputed patent, a method of matching different response variable values is classified into at least two categories based on 1) the number of disputed claims of the disputed patent as a response variable value, and 2) the number of disputed claims of the disputed patent, N differentiation methods such as the method of using the category value as a response variable value (e.g., dividing the dispute into one or two or more disputes, or by dividing into four quartiles by considering the distribution of disputes by dispute patent). 3) A matrix of n * m cells is formed by considering both the n-quartile based on the number of disputes and the m-quartile based on the disputed year of disputed patents (the number of disputed patents in each cell should be as equal as possible). It may be a method of mapping the response variable value to each cell. On the other hand, the disputed patents related to two or more of the disputed patents can be adopted as a method of increasing the number of disputes, or a method of matching different response variable values depending on the level of the dispute as independent disputes by class.

Meanwhile, among the broad dispute patents, there may be a sub-dispute patent. Quasi-dispute patents are patents in accordance with disputes, which include: 1) Class I semi-dispute patents relating to non-judicial administrative disputes, such as invalid trials or ITC complaints; Quasi-dispute patents, 3) royalty-patented patents, third-class quasi-dispute patents subject to cross licensing, and 4) other quasi-dispute patents for exercising legal rights. The quasi-dispute patent related to 1) is relatively easy to collect data by the Patent Office or other methods, while 2) to 4) occurred in a private area, which is difficult to collect data. However, 2) to 4), when creating a conflict prediction model for each user (e.g., specialized for A for a particular company A), each user can know 2) to 4) associated with him, To 4) may be used to generate a dispute prediction model. 2) to 4) can be importantly used for generating a company-specific dispute prediction model.

The quasi-dispute patent may also have a method of matching a response variable value identical to a dispute patent (patent in which a judicial dispute has occurred) and a method of matching a response patent value with another response variable value. The latter is more preferable. As the simplest example of response value response, there may be a method of allocating 1 for both disputed and semi-disputed patents and 0 for non-disputed patents.

The dispute prediction model generation unit 5520, when the dispute prediction model value corresponding to the dispute prediction model value and the response variable value corresponding to the explanatory variable value for the dispute patent and non-dispute patent, as described above, the explanatory variable value and the response variable value A conflict prediction model is generated by applying at least one statistical modeling method that is set in advance.

If the dispute patent and the semi-dispute patent all correspond to 1 as the response variable value and all the non-dispute patents correspond to 0, generating the dispute prediction model is to generate a typical classification model. The method for generating a classification model by the dispute prediction model generator 5520 may employ various statistical techniques, all of which may be employed in the present invention, and are included in the statistical processing method of the present invention. It will be clear that the lack of a detailed description of the statistical technique does not exclude the present invention.

In this specification, as an example of generating a classification model, a method of generating a classification model using boosting among ensemble techniques based on a machine learning technique tree will be described in detail. Another method using the ensemble technique is the random forest technique. FIG. 27 is a diagram illustrating a summary of the Gradient Boosting algorithm algorithm, and FIG. 28 is a summary of an algorithm proposed by Friedman (2002) for a newly stochastic gradient boosting algorithm. 29 is a view for explaining an exemplary concept of a dispute prediction model generation process of the patent dispute prediction information generation system 5000 of the present invention.

 Next, the information processing method of the dispute prediction model generation part 5520 of this invention is demonstrated in detail, referring FIG. 29 and FIG. The dispute prediction model generation unit 5520 generates a dispute prediction element value for each dispute prediction element, which is an explanatory variable, for dispute patents and non-dispute patents, and sets a dispute patent response response variable value for the dispute patent, Non-dispute patent-response response variable values are set (SLBO12). Subsequently, the dispute prediction model generator 5520 generates a first stump for any one or more of the explanatory variables (SLBO13), and generates a dispute prediction model value using a first set of stumps configured with the generated first stump ( SLBO14) and the dispute prediction model value is verified (SLBO15). If the verification result does not meet the preset criteria (SLBO16), after determining the tree importance, adjust the misclassification patent target weight (SLBO17), generate a second stump for the explanatory variable, and use the first stump and the second second stump. Generate the dispute prediction model value, verify the dispute prediction model value, and generate the stump until the verification result meets the preset criteria, and generate the dispute prediction model value and verify the dispute prediction model value using the generated stumps. Repeat. If the verification result meets the preset criteria, the generated stumps are then joined to determine the dispute prediction model (SLBO18).

The dispute prediction model generator 5520 generates a stump for each explanatory variable. In some cases, you can use generic trees in addition to stumps. The dispute prediction model generation unit 5520 generates at least one stump including an important dispute prediction element by analyzing the values of explanatory variables of the dispute patent constituting the dispute patent set and the non-dispute patent constituting the non-dispute patent set. . The stump includes five pieces of information, such as 1) conflict prediction factor, 2) split point, 3) left node prediction value, 4) right node prediction value, and 5) prediction when branch is not applicable. The rule for generating branches preferably seeks to find branch origins and prediction values to minimize a given loss function. In this case, a loss function or a loss function can be used as a loss function.

32 shows an example of the generated stump, and Table 5 below shows an example of the stump information.

Splitvar SplitCodePred Leftnode RightNode Missingnode ErrorReduction Weight Prediction 0 11 0.5 One 2 3 802.9285468 22666 0.000128321 One -One 0.007353469 -One -One -One 0 13327 0.007353469 2 -One -0.010182157 -One -One -One 0 9339 -0.010182157 3 -One 0.000128321 -One -One -One 0 22666 0.000128321

In Table 5, SplitVar refers to a branching explanatory variable (disput prediction element), and a value of 11 refers to the eleventh explanatory variable. SplitCodePred is the branch point where the explanatory variable is branched, and 0.5 is the branch point value. LeftNode represents the left node, where 1 corresponds, and the third row starting with 1 gives information about the left branch time. In the left branch, X11 <= 0.5, the predicted value in this case is 0.007353469, and 0.007353 is shown in FIG. RightNode is the right node and the information value in this case is shown in the fourth row, and MissingNode is for branching when missing. The information value in this case is shown in the fifth row.

Table 6 below shows an example of generating a plurality of stumps, including the information of Table 5.

Splitvar SplitCodePred Leftnode RightNode Missingnode ErrorReduction Weight 11 0.5 One 2 3 802.9285 22666 -One 0.007353 -One -One -One 0 13327 -One -0.01018 -One -One -One 0 9339 -One 0.000128 -One -One -One 0 22666 11 0.5 One 2 3 773.5689 22666 -One 0.007046 -One -One -One 0 13238 -One -0.01016 -One -One -One 0 9428 -One -0.00011 -One -One -One 0 22666 11 0.5 One 2 3 785.034 22666 -One 0.007055 -One -One -One 0 13242 -One -0.0103 -One -One -One 0 9424 -One -0.00016 -One -One -One 0 22666

As can be seen in Table 6, four rows correspond to one stump. Table 6 shows that at least three stumps are generated for the explanatory variable 11. As described above, one or more stumps are generated for one explanatory variable.

The dispute prediction model generation unit 5520 may apply the first stump S1 and use the dispute prediction element value of the dispute patent and the dispute prediction element value of the non-dispute patent to set the dispute patent of the dispute patent set and the non-dispute patent of the non-dispute patent set. The prediction value S1 is applied to any one of the 3) left node prediction value, the 4) right node prediction value, and 5) the prediction value when the branch is not applicable. The first stump S1 becomes the first dispute prediction model candidate. Prediction values will be generated for each disputed and non-disputed patent used to generate the model. In this case, the dispute prediction model generation unit 5520 analyzes the response variable value for the dispute patent, the prediction value for the dispute patent, the response variable value for the non-dispute patent, and the prediction value for the non-dispute patent, such as a misclassification rate. Produce predictive error information. The dispute prediction model generation unit 5520 determines a method of increasing the weight (forcibly) of weights (for example, frequency) for misclassified dispute patents and misclassified non-disputed patents after determining the importance of the tree. For example, if a specific patent Pi is a disputed patent and is classified as a non-disputed patent, the frequency of Pi is increased from 1 to more), and the disputed patent set data and the non-disputed patent set data are readjusted. A second stump S2 is generated for the readjusted disputed patent set and the non-disputed patent set. The second dispute prediction model candidate consists of S1 and S2. At this time, the dispute prediction model generation unit 5520 generates a prediction value of the dispute patent and non-dispute patent by applying S1rhk S2. Regarding the disputed patent and the non-disputed patent, a comparison between the response variable value of the disputed patent and the response variable value of the non-disputed patent and a dispute prediction model value which is a prediction value generated by the second dispute prediction model candidate consisting of S1 and S2 Generate prediction error information. The dispute patent and non-dispute patent data are readjusted in consideration of the generated prediction error information. As described above, each patent Pi constituting the disputed patent set and the non-disputed patent set is composed of the i-th stump set (SSi, ith Stump Set, a set consisting of the stumps of S1 to Si, and the candidate for the i-th dispute prediction model). to generate a predicted value, and by comparing the response variable value of the generated prediction value and Pi generating the error prediction information, and with reference to the generated error prediction information readjust the dispute Patent data set and non-conflict Patent data set S _{i +} generating a _first and a (i + 1) th candidate conflict prediction model using the (i + 1) th set are generated stump.

Through the above process, a stump set composed of significantly large n stumps S1, S2, and Sn is generated. For each of the stumps Si, 1) a dispute prediction element, 2) a split point, 3) a left node prediction value, 4) a right node prediction value, and 5) a prediction value when branch inapplicability is applied. 1) to 5) for each of n Sis are candidates for dispute prediction model.

If there is a dispute prediction model candidate, the method of using the dispute prediction model candidate is as follows. The dispute prediction model generation unit 5520 inputs a dispute prediction element value for each dispute prediction element of Pi into the dispute prediction model candidate for any patent Pi. The dispute prediction model generation unit 5520 may determine a dispute prediction element value for each dispute prediction element at least one or more stumps (in a very special case, there may be no corresponding stump, in which case the dispute prediction element is not used in the model). To generate a prediction value for each stump. The predicted value is any one of the 3) left node predicted value, the 4) right node predicted value, and the 5) branch predictable value. When the generated prediction value is summed for all the stumps, a dispute prediction model value is generated. That is, the dispute prediction model generator 5520 may add a dispute prediction model value for each Pi in such a manner as to add all the predicted values generated by applying the dispute prediction element values for each dispute prediction element of Pi to the stump set which is the dispute prediction element candidate. Create When the dispute patent response variable value is 1 and the non-dispute patent response variable value is 0, the dispute prediction model value for each Pi becomes a value between 0 and 1. At this time, if the dispute prediction model value of Pi is less than 0.5 on the basis of 0.5, Pi may be predicted as a non-dispute patent, and if it is 0.5 or more, it may be classified as predicting a dispute patent. On the other hand, since Pi is a disputed patent or a non-dispute patent, the response variable value corresponding to Pi becomes a true value, and this true value and a predicted value are generally different.

Next, a method of generating a dispute prediction model having the optimum or validity by the dispute prediction model generator 5520 will be described. When there are candidate candidates for the conflict prediction model using n stumps, the generation of the conflict prediction model is generated through cross validation. In general, the larger n, the higher the predictive performance of a train set (consisting of disputed patents and non-disputed patents) that generate candidates for conflict prediction models.However, excessively large n results in overfitting, resulting in a test set. If the second dispute patent set and the divided second non-dispute patent set are used to verify the generated dispute prediction model candidates, they may be test sets. Therefore, it is important to determine n of the appropriate size. The dispute prediction model generation unit 5520 generates a dispute prediction model value Mi (Pi) for each patent Pi included in the test set with respect to the generated dispute prediction model candidates M1, M2, and Mi. Mi (Pi) refers to a prediction value generated by applying a dispute prediction element value for each dispute prediction element of Pi to a dispute prediction model candidate Mi. On the other hand, the reaction variable value of Pi is referred to as Y (Pi). When the response variable values are 0 and 1, Y (Pi) is 0 or 1. Subsequently, the dispute prediction model generating unit 5520 obtains Y (Pi) and Mi (Pi) values for all patents Pi constituting the test set, and substitutes these values into a given loss function to minimize Mi. Select.

30 illustrates a conceptual diagram in which the concept of overfit is expressed. In the generation of a model that distinguishes a red dot from a blue dot in FIG. 30, if a model such as a green line is generated, the train set is distinguished very well. However, in this case, the predictive performance of the test set does not participate in the model generation. There is a problem of falling. At this time, if a model such as a black line is generated, it becomes an appropriate classification model.

Next, the shrinkage parameter will be described. The tree importance determination algorithm is used to determine the importance of the tree. The performance of the model is improved by introducing the shrinkage parameter into the prediction model such as multiplying the tree importance determination value derived from the tree importance determination algorithm by the shrinkage parameter. In the tree importance determination algorithm, tree importance is determined by performing a cross validation process while varying branch points for each conflict prediction element.

When the dispute prediction model generation unit 5520 generates the dispute prediction model, if the size of the dispute patent set is large enough (the majority of the patents do not have disputes among all patents, the size of the non-dispute patent set is sufficient). The dispute patent set may be divided. However, if the size is not sufficient, the n-fold cross validation technique may be used. 33 illustrates a 5-fold cross validation scheme. The dispute prediction model generation unit 5520 divides the data containing the disputed patent and the non-dispute patent into five sets, and then builds a model with four sets in each case and obtains a test error with the other set. N-fold Cross validation method can be implemented as the method of selecting the smallest model.

In the above, the classification model for the two categories has been described. Meanwhile, the classification model method described above may be applied to n categories (non-dispute, dispute, multi-dispute, etc.). Disputes Patents can be classified as disputes if they have a range of 1 to a predetermined value, and if they exceed a predetermined value, they can be classified as multiple disputes. In this case, the response variable value may be 0 for non-dispute, 1 for dispute, and 2 for multidispute. In this case, 2 is a category distinguished by 2 as a numerical meaning, and it may be more appropriate to treat it as a non-dispute or a classification group having an attribute different from the dispute, and 3 or another numerical value other than 2 responds. It could be a variable value. In this case, the response variable may be expressed as a vector. For example, data with (1, 0, 0) represents data in which the response variable is in the first category, and data with (0, 1, 0) represents data in the second category.

Next, referring to FIG. 34, a method of generating the dispute prediction model using n classification models as a method of generating a classification model for three categories A, B, and C will be described. Unlike in FIG. 32, when there are three or more categories, a stump is generated for each category. That is, a stump for category A, a stump for category B, and a stump for category C are generated, respectively. Then, from each stump, we calculate the probability of belonging to that category and assign each piece of data to the category with the highest probability. For example, if one particular data is assigned to three stumps, and each of the predictions is -0.567, 0.456, and 0.234, the predictions for category B are the highest, so this data is most likely in category B. Therefore, this data is classified as category B.

Next, a regression model will be described with reference to FIG. 35. The regression model is a statistical model used when the response values are continuous. The dispute patent can be divided into semi-dispute patents and dispute patents, and it can be assumed that semi-dispute patents have a predetermined importance (for example, 70%) relative to dispute patents in relation to disputes. In this case, a response variable value of 0 may be assigned to the non-dispute patent, 0.7 to the semi-dispute patent, and 1 to the dispute patent. The values of 0 may be assigned to the non-disputed patent, 1.4 to the semi-disputed patent, and 2 may be assigned to the disputed patent 2, respectively. There will be. On the other hand, for multi-dispute patents, one may be assigned to each dispute. In this case, 0, 1-time dispute patents 1, 2-time dispute patents 2, .., n-time dispute dispute patents We can assign the response variable n. On the other hand, the number of disputes is divided into n categories, and for each category, the predetermined importance value for each dispute-patented patent (2 to 5 times, 3 to 6 to 20 times, 3 to 20 times, etc.) If assigned, non-disputed patents have 0, 1 disputed patent 1, 2-5 disputed patents, 2, 6-20 disputed patents 3, 20 disputed patents 3 and 3 You can assign it by value. On the other hand, a semi-disputed patent may be assigned a value equal to 0.7, and if the importance is handled differently according to the type of the semi-disputed patent, different response variable values may be assigned to the types of the semi-disputed patent.

A method of generating a dispute prediction model from a regression model may also use the same stump as used in the classification problem as shown in FIG. 32. However, the loss function used here uses the least square error derived from the normal distribution, not the deviance function derived from the likelihood function of the binomial distribution. And the meaning of the predicted value is also different. The predicted value in the classification problem is related to the probability that the response variable value of the data is 1, but the predicted value in the regression model is the predicted value of the response variable itself. In Fig. 35, for the explanatory variable X3, it can be seen that the value of X3 is branched based on 18.6, and the predicted value for each branch is displayed.

How to make stumps in regression problems is no different from classification problems. Only the loss function used is changed and the optimal stump is generated according to the changed loss function. In the classification problem, we use the loss function based on the likelihood function of the binomial distribution. In the regression problem, we use the least square error as the loss function. The meaning of the predicted value is also different. The predicted value in the regression problem predicts the response variable itself, and the predicted value in the classification problem is related to the probability that the response variable has a value of 1.

On the other hand, as a special case of the regression model, it will be possible to generate a re-dispute prediction model among dispute patents targeting only dispute patents. In this case, the sample consists only of the disputed patent, and the response variable value is 1) the number of disputed claims of the disputed patent, or 2) the number of disputed claims, or 3) the category of the disputed claims, or 4) the category of the number of disputed claims. Value. 1), 1 dispute dispute patent is 1, 2 dispute dispute patent is 2, n dispute dispute patent can be assigned as a response variable value n, 2) 1 dispute dispute patent is 0, 2 dispute dispute Patents for 1 and n disputed patents can be assigned a response variable value of n-1, and for example, 3), one disputed patent is 1, 2 to 5 disputed patents are 2, For 6-20 disputed patents, dispute disputed patents exceeding 3, 20 can be assigned as a 4-valued response variable.For example, 4) disputed patents have 0, 2-5 disputes. 1, 6-20 disputed patents may be assigned to 2 or 20 disputed patents, and 3 may be assigned as a response variable.

The dispute prediction model generating unit 5520 is the same method as the information processing method for generating the dispute prediction model with the dispute prediction element value for each dispute prediction element of the dispute patent Pi constituting the dispute patent set using the response variable values as described above. Create a re-dispute prediction model. The value generated by the re-dispute prediction model becomes a re-dispute prediction model value. The reason for this re-dispute prediction model may be used to determine whether a patent of interest will cause re-dispute or how many disputes it will cause.

 On the other hand, the dispute prediction model generation unit 5520 described the generation of a dispute prediction model focusing on the occurrence of disputes or the number of occurrences of disputes. Next, a method of generating a dispute point prediction model centered on the dispute point of time of the dispute prediction model generation unit 5520 will be described. Prediction of a dispute is a prediction of when a dispute is raised by the value of a dispute prediction element of a specific patent Pi.

Next, with reference to FIG. 36, the dispute prediction model value generation part 5530 for each patent of this invention is demonstrated. The dispute prediction model value generation unit 5530 for each patent obtains a patent for generating a dispute prediction model value (SL41), generates or obtains a dispute prediction element value of the obtained patent (SL42), and sets a dispute prediction element for a patent. The value is input to the dispute prediction model to generate a dispute prediction model value of the patent (SL43), and the dispute prediction model value storage of the generated patent (SL44).

The dispute prediction model value generating unit 5530 for each patent performs a function of generating a dispute prediction model value by applying a dispute prediction model to Pi. The dispute prediction model value generation unit 5530 for each patent may also function in generation of a dispute prediction model (the dispute prediction model generation unit 5520 also applies a dispute prediction model candidate Pi to a dispute for a dispute prediction model candidate. The generation of prediction model values has been described. In this case, the patent-specific dispute prediction model value generation unit 5530 may operate under the dispute prediction model generation unit 5520 or the dispute prediction model generation unit 5520. May be used by calling the function of the dispute prediction model value generation unit 5530 for each patent.) The dispute prediction model generated by the dispute prediction model generation unit 5520 is used independently of the dispute prediction model generation unit 5520. Can be used to generate a dispute prediction model value for each input patent Pi. The dispute prediction model value generation unit 5530 for each patent determines a dispute prediction element value for each dispute prediction element generated by the dispute prediction element value generation unit 5510 with respect to patent Pi (SSn composed of a set of n stumps). Is an example of a dispute prediction model, which is also applied to a dispute prediction model generated using other statistical processing techniques, to generate a dispute prediction model value of Pi.

The dispute prediction model value may be a value that has undergone a predetermined post-processing. Next, the post-processing will be described. It does not matter if the data used in constructing the model (disputed patent set and non-disputed patent set) are randomly extracted from the entire population, otherwise statistical post-processing is required. Since there are only tens of thousands of disputed patents in the population (about 4 million US patents), using a random sample (disputed patent set + non-disputed patent set) in these populations results in a very small sample size. The very small number of disputed patents in a sample makes it difficult to create a good model. As many dispute patents are included in the sample as possible and analyzed to understand the characteristics, a good model can be constructed, and the data are analyzed by the case-control study method. In the case of patent data, information on the disputed patent has already been collected, so only the non-disputed patent data corresponding to the patent can be randomly extracted and analyzed by the case-control study method. It is important to note that the score obtained from the case-control sample is not what the original population wants to know, so a procedure is required to reduce the original score. In this case, the size of the entire population must be known. For example, in the case of US patent data, because we know the size of the entire population (about 4 million), we can easily get the original score using Bayes' theorem. The dispute prediction model value generating unit 5530 for each patent may obtain a final dispute prediction model value (final score) using Equation 1 derived from Bayes' theorem. In Equation 1, n1 denotes the number of disputed patents in the sample, n0 denotes the number of non-disputed patents in the sample, N1 denotes the total number of disputed patents, and N0 denotes the total number of non-disputed patents. The boosting score is a dispute prediction model value obtained by applying a dispute prediction element value corresponding to Pi to the dispute prediction model generated by the dispute prediction model generator 5520.

[Equation 1]

Equation 1 is a method of obtaining a final score when the control sample is randomly extracted, and the equation is different when the control sample is stratified. The size of each category is different and the statistical properties of the dispute prediction model value are different for each category based on each patent classification.To properly reflect this, the non-dispute patents are stratified and extracted to correspond to the patent classification of the disputed patents. Is considered. Stratified extraction is used to ensure that samples are evenly sampled for each patent classification category. When the sample is generated by stratification extraction, the equation for obtaining the final score is shown in Equation 2 below. Where s is an index identifying a patent classification category. For example, n _1s represents the number of disputed patents sampled from the sth patent category.

&Quot; (2) &quot;

As described above, the dispute prediction model value generating unit 5530 for each patent generates a dispute prediction model value for each patent. Meanwhile, the dispute prediction model value generation unit 5530 for each patent may generate the dispute prediction model value for all patents in the patent DB 1120, and the patent dispute prediction information generation system 5000 or the The dispute prediction model value may be generated for a patent composed of at least one patent designated or managed by a user of the patent dispute prediction information generation system 5000. The generated dispute prediction model value is stored in the dispute prediction model value DB 5230. The dispute prediction model value providing unit 5540 of the present invention performs a function of providing a dispute prediction model value for the patent when receiving a request for dispute prediction model value for at least one or more patents from the outside. When the user selects at least one or more patents or the list of patents of at least one director is presented as a search result, the dispute prediction model value providing unit 5540 may select the selected patent. Alternatively, the dispute prediction model value may be provided to include the dispute prediction model value in the search result.

Next, a method of generating dispute prediction information through the dispute prediction model will be described with reference to the drawings.

The dispute prediction engine 5100 first obtains a self patent set including at least one patent and obtains at least one target patent set having a predetermined relationship with the self set, or obtains a target patent set including at least one patent. First, at least one self patent set having a predetermined relationship with a target set is obtained (SL51). Subsequently, the dispute prediction engine 5100 obtains at least one dispute prediction model value for each patent for each patent constituting the target patent set (SL52), and utilizes the dispute prediction model value for each patent to at least one or more patents. The dispute prediction information is generated (SL53). Separate it below.

First, a self-patent set (SS) is defined. The self patent set refers to a patent set specified by a patent group including at least one patent corresponding to a preset input of a system or an input input by a user for generation of dispute prediction information. The input may be any one or combination of at least one patent input, at least one owner input, at least one inventor input, at least one patent classification input, and at least one search input. The search expression naturally includes a search expression for each field and field combination for all fields capable of calling a patent stored on the patent DB 1120, and naturally includes an input of a search keyword for a specific field. By way of example, a user may apply to a company's technology tree, a domestic or foreign or country-specific competitor or affiliated company tree, a technology field tree corresponding to a function related to his product, or at least one director's patent group. A search tree that can call a patent group of interest, and a patent group of interest, and end nodes of each tree may correspond to user inputs or conditions or individual patent documents. It will be appreciated that the tree can be a multi-stage tree of at least one hierarchy. The IPC hierarchy is a representative example of the multistage tree. Generating the SS is performed by the self-patent set generating unit 5110 of the present invention.

The self-patent set is a patent set composed of at least one published or registered patent generated, selected, or designated by the user or the patent dispute prediction information generation system 5000. [0047] Self-patent set generation of the dispute prediction engine 5100. The unit 5110 generates or obtains a self-patent composed of at least one patent to form a self-patent set. The generation of the self patent set may be generated through a search through a search engine or a DBMS, and designation of the at least one patent set stored by a user or a patent set generated as a result of the operation of the patent set or the patent set. It may also be generated through designation of at least one patent included in. The self-patent set does not need to be a patent owned by the user's organization, but a patent that the user designates or selects with interest is sufficient. The user may manage patents of interest in a tree of a multi-level hierarchy. Each node constituting the tree has a node name, and at least one patent is associated with each node.

The reason for introducing the SSi is as follows. A patent dispute is in essence a matter of the relationship between the plaintiff's patent rights and the defendant's machine, product, method, and composition (hereinafter referred to as the product) and not the plaintiff's patent and the defendant's patent. Accordingly, it is necessary to project or project the defendant's product into data that the system can process. Accordingly, the user of the system needs to input the patent group related to the product of interest so that the system can understand. Examples of the input of patent group related to the product include 1) the input of patents if there is a patented technology among the technologies reflected in the product, such as the function, structure, method and material of the product, and 2) no or insufficient patent of the company. Search for retrieval of third-party patents related to our products, 3) patents of others related to our products, 4) entering patent classifications in technical fields related to our products, and 5) patents related to our products. It is possible to specify a patent group associated with a product through the input of any one or more or any combination of one or more of the inputs of the equation. The system will generate patent dispute prediction information from a group of patents on which a product is projected, not a direct product.

When the user sets the SS, a subjective weight may be set for the SSi. The setting of the weight is performed through a UI provided by the weight adjusting unit 5153 of the present invention. The UI provides a list of patents belonging to the SS, and the user can enter a weight, importance or importance level in the patent list. This weight is referred to as user set SSi weight. The user setting weight is also reflected in a later-described TS generated by the SSi.

Let SSi be a patent belonging to its own patent set SS. Define a target patent set (TS) for the SSi individual patent. TS refers to a set of patents specified by a user or a system generated, designated, or inputted for the purpose of generating SSi and dispute prediction information. Generating the TS is performed by the target patent set generation unit 5120 of the present invention. The target patent set is a patent set including a target patent which is a patent having a predetermined relationship with the own patent included in the own patent set. The target patent set generation unit 5120 of the present invention generates a target patent set through its own patent set, or generates a target patent set through designation or selection of the user or the patent dispute prediction information generation system 5000. The former will be described in detail. The preset relations include 1) citation-cited patent relations, 2) similar patent group relations in text mining techniques, 3) similar technology patent relations in patent classifications, and 4) designation patent group relations consisting of user-specified patents. This can be.

Next, a method of generating dispute prediction information by using the dispute prediction engine 5100 using weight information for each target patent in consideration of predetermined relationship information will be described. The dispute prediction engine 5100 obtains a self patent set (SL61), extracts a target patent having a predetermined relationship for each individual patent constituting the self patent set, and generates a target patent set (SL62). Subsequently, the dispute prediction engine 5100 obtains or generates relationship information about the target patent (SL63), and then generates weight information for each target patent in consideration of the relationship information (SL64), and then for each target patent. By using the weight information, at least one preset dispute prediction information is generated (SL65).

FIG. 37 shows that a predetermined relationship is established between the self patents constituting the self patent set and the target patents constituting the target patent set. The relationship is defined as R (SSi, TSi), which is a value established between SSi and TSi. As can be seen in FIG. 37, one SSi may be associated with at least one TSi, and one TSi may be associated with at least one SSi. In conduit 1, R (SS1, TS1), R (SS1, TS2) and R (SS2, TS2) may be different, and in terms of TS2, two relations R (SS1, TS2) and R (SS2, TS2) ) The TSi may have n relationships, and the relationship may be used to generate W (TSi), which is a weight for TSi. For example, when the relationship is a similar patent group relationship in text mining, R (SSi, TSi) may be a degree of similarity to which a preset keyword similarity function based on a key keyword between SSi and TSi is applied. On the other hand, if the relationship is a citation, R (SSi, TSi) may be a citation relationship function value added considering the type of citation of SSi and TSi.

The citation-cited relationship is a self-patent and a target patent having at least one relationship of 1) direct citation, 2) citation depth n (n> 1) indirect citation, 3) latent citation, and 4) chain citation. As described above. If there is an individual patent Pi, a different citation weight preset for each citation type may be provided. The similar patent group relationship in the text mining technique has the keyword similarity set based on the keyword from which the own patent and the target patent are extracted, and the similar technology patent relationship in the classification is the similar technology that the own patent and the target patent are preset in the patent classification system. To have a relationship.

The cited patent set generation unit 5121 of the present invention generates a cited patent set from the preceding patents cited by the Pi for each Pi cited type, for the Pi belonging to its own patent set, and the cited patent set is a target patent. An example of a set or a subset of a target patent set. When generating the cited patent set, the Pj patent belonging to the own patent set may be included in the cited patent set. In this case, whether to include Pj in the cited patent set depends on the target patent set generation policy. On the other hand, when the target patent set generation unit 5120 generates the target patent set, the user is a specific applicant (for example, if the user belongs to company A, applicants having a specific relationship with company A or company A, etc. Patents having predetermined attributes such as patents) may be excluded. Of course, the exclusion of the patent of the particular applicant may be processed through the post-processing (deleting, adding, etc.) of the user to the generated target patent set.

At this time, TSi, which is a patent belonging to TS, which is a cited patent set, may have TSi having different weights. The cited patents that make up the cited patent set have 1) citation weights based on citation type, 2) weight citations based on duplicates, 3) citation depth based citation weights for indirect citations, 4) inverse citation counts, and 5) citation obsolescence weights. . Quotation type-based citation weighting can be given in such a way that the direct weight is given the largest weight, different weights are given for each detail type of potential citation, and in the case of indirect citation, the weight is lower as the citation depth is larger. Could be. 4) The inverse weight of the citation count is that the larger the number of references in SSi, the lower the relative weight of a specific reference. There has been a recent abuse of references (adding as many patents and articles as possible), and if the number of references is too high, it is likely that there are references that do not actually act as references. The quoted inverse weight may be "constant / (reference number-preset integer)", which is applied only when the reference number exceeds a preset integer. As the date of filing between SSi and TSi increases, there is a high possibility that the relevance of content may be reduced even if the same patent classification is used. Therefore, the cited obsolescence weight may be added to TSi in consideration of an obsolescence function having an inverse relationship to the difference between the application date of TSi and SSi. The citation obsolescence weight may be "constant / f (the number of days of application of SSi and the date of application of TSi)"

On the other hand, since n different SSi may quote the same TSi, in this case, the weight according to the duplicate citation of TSi is n. On the other hand, SSi cannot directly treat the parent patent of citation depth 1 and the grand parent patent of citation depth n (n> 1) cited by the parent as equal weights. At this time, as the citation depth n increases, a weighting factor of TSi according to the citation depth is adjusted by adding a decay factor that decays exponentially, exponentially, or arithmetic exponentially. On the other hand, additional weights are assigned in the case of the recently cited among TSi, the weights are further assigned according to the property of the owner, such as when the owner is a registered competitor of a troll, multi-dispute or other user, or standard patent pool If the patent has a predetermined attribute, such as a patent belonging to or a disputed patent, it may be given a weight more. The weighting of the TS by the system is performed by the weight adjusting unit 5103 of the present invention.

The weight based on SSi is demonstrated. The user may assign a user set SSi weight which is a subjective weight for each SSi patent set by the user. Since TSi generated by cited patent sets or similar patent groups are based on SSi, weights assigned to SSi can be added to TSi. That is, TSi, the patent group cited by a particular SSi having a very high weight, should have a higher weight than TSj, a patent group cited by a low SSj. Therefore, it is more preferable that user set SSi weight information, which is a weight based on SSi, is added to TSi constituting the TS.

Another representative example of the TS is a similar patent group generated through a text mining technique for patents belonging to the SSi. This will be described. At least one keyword candidate is first extracted for all individual patents, and then a key keyword is extracted and stored from the keyword candidates, and then at least one or more patents having a key keyword with high similarity to the key keyword group corresponding to any patent i. To generate a similar patent group. The keyword candidate includes a keyword pair (air pair) and a patent classification code, and may be processed like a keyword. The generation of the similar patent group is in charge of the similar patent group generation unit 5122 of the present invention. The similar patent group generation unit 5122 extracts a similar patent application group similar to the SSi from the key keyword set extracted for the SSi as the material. Generate the SSi and the similar patent group TS (SSi). The SSi and the TSi (SSi) has a similarity score. On the other hand, TS (SSj) is also generated for the other patents SSj constituting the SS, and TSi constituting the TS may have the same TSi as the similar patent for both SSj corresponding to at least one or more j instead of SSi and i. In this case, the frequency of TSi may be 2 or more.) And the same number of similarities as the frequency.

In particular, in the case of patent classification, each of the plurality of patent classifications may be used as a key keyword. On the other hand, it is possible to give a different weight for each patent classification in a manner that further weights the main patent classification. On the other hand, the deeper n deep in the patent classification system, such as n deep sub n dot subclass of the patent classification, the deeper patent classification can be assigned a higher weight. Meanwhile, when there is a patent classification of H04B 7/26, H04B 7/24 (1dot Subgroup), H04B 7/00 (Main Group), etc., which are parents of H04B 7/26 (2 dot Subgroup), are not specified. , Because it represents the technology on the patent specification, it can be treated like a key keyword. Of course, the weight for H04B 7/26 is high, H04B 7/24 is low, and H04B 7/00 (Main Group) may be set to the lowest. Furthermore, when generating similar patent groups by text mining techniques, reference information (patent numbers related to direct citations, indirect citations, and potential citations) may be treated as key keywords and used in generating similar patent groups. In this case, the same or different weights may be given for each citation type and citation depth for each citation relationship patent.

The method of generating a similar patent group includes a clustering method and a method using a search engine. Clustering is a method of generating a similar patent group based on the distance between key keywords. The simplest method is the K-Means algorithm. On the other hand, since the search engine includes a ranking algorithm in itself, when there are n key keywords input, the search engine outputs a patent document including a plurality of n key keywords among the searched patent documents. Patent groups that become a certain number or a predetermined score or more can be treated as a similar patent group. At this time, each of the key keywords may be processed to have a different weight for each field in the patent specification (for example, to give the highest weight to the key keywords in the name of the invention, the key keywords in the claims For example, if the search engine is able to obtain the weight for each keyword, then the n cores are entered. You can query the search engine by weighting the keywords. In particular, if a query is generated with n key keywords including weights, and the generated query is queried into a key keyword field (a field collecting only extracted key keywords), a similar patent group may be rapidly generated.

Next, a method of generating key keywords will be described. The keyword may be generated by processing the text included in the patent specification, which is in charge of the key keyword generator 2100 of the data processor 2000 of the present invention. The key keyword generator 2100 extracts a keyword from a phrase or sentence corresponding to each field of the patent specification. The extraction of the air pair is extracted through the combination between the terms in the proximity distance (the distance satisfying the distance standard between the registered terms in one sentence). The field may be any one or more of various fields constituting the patent specification, such as the name of the invention, the claims, the summary, the detailed description of the invention, industrial applicability, effects, and the prior art (background art). The key keyword generator 2100 generates a key keyword set for the n keywords extracted from the field. When generating the key keyword, the key keyword generator 2100 performs synonym processing, thesaurus processing, and the like to select the key keywords by grouping the terms having substantially the same, equivalent, or equivalent meaning as a representative term. It is preferable. Meanwhile, when processing with the representative term, it may be desirable to perform synonyms and thesaurus processing for two or more languages in one patent document by using a dictionary or a machine translator. On the other hand, it is more preferable that the representative term or the extracted key keywords are translated into at least one language through the dictionary or machine translator. Extracting a keyword (commonly referred to in this technical field as a term) or an air pair through natural language processing belongs to techniques known in the field of natural language processing technology. For the extracted n keywords (including air pairs, it is natural), a core keyword set (a core air pair set is selectively included to represent the patent specification by applying a predetermined key keyword selection algorithm) Will be selected). The most frequently used algorithms use term frequency (TF) and inverse document frequency (IDF).

In the field of natural language processing technology, various functional expressions using TF and IDF as variables are disclosed, and it is natural that other complex equations such as weights for each field can be applied in a policy manner. At this time, the core keyword set composed of only the narrow meaning keywords and the key keyword set composed only of the air pairs are generated separately or the keywords having the narrow meaning and the air pair are equally processed by the algorithm. When the key keyword selection algorithm processes the air pair together with the narrow key keyword, a key keyword pair set may be generated for n keyword pairs.

Equation 3 below is an example of an equation used in an algorithm for extracting a key keyword with respect to the extracted keyword.

&Quot; (3) &quot;

tf: term frequency The frequency at which the keyword (index) appears in the current document.

N: the total number of documents

df: the number of documents in which the keyword appears in document frequency

slope: slope (arbitrary value, adjustable)

ut: unique terms for the entire document set

pivot: average length of literature (ut / N)

uf: ut of the document

Fwi: weight by field

On the other hand, when the core keyword generator 2100 extracts a new term consisting of two or more words / vocabulary / words, it is often difficult to determine whether the new term is a term having technical meaning. have. In this case, there may be a method of determining whether a new term has a technical meaning by using an external search engine such as google.com. The core keyword generation unit 2100 performs at least one or more predetermined processing such as quotation processing (queries processing method of google.com that processes an exact match) and the like, and then extracts the new term. When the search result is transmitted to an external search service system, the search result is received from the external search service system, and the search result is satisfied, a predetermined criterion is satisfied, and the extracted new term is processed as a normal term. The analysis of the search results measures the number of search results (the number of hits, which teaches how many search results match the query), and examples of the preset criteria are 1,000 or more in English and other languages. It could be more than 100 and so on. For example, as of March 2010, querying google.com for "patent informatics" and "patent informatics services", respectively, yields 67,300 and 279 results, in which case "patent informatics" is treated as a term. patent informatics services "may not be treated as a new term. On the other hand, you can query a system that provides a description of a term, such as wikipedia.org, rather than a search engine such as google.com, and treat the term as a new term if it exists.

Through the above-described key keyword selection algorithm, it will be obvious that one or more key keyword sets corresponding to one patent document may be selectively included. That is, the key keyword set KS (Pi) = {K1 (Pi), K2 (Pi), ,,,, Ki (Pi), Kj (Pi) ,,,, Kn (Pi) corresponding to the i th patent document Pi } You get In the above description, i, j, n are integers, and Kn (Pi) refers to the nth key keyword selected from the i th patent document Pi. A set of key keywords can be mapped to a single patent document, because: 1) specific fields (for example, claims or summaries), 2) different weights for each field, 3) selection of two or more key keywords There may be various reasons, such as the use of an algorithm, 4) reference range of IDF calculation, 5) term extraction method, etc.) Core keyword set KS (Pi) = {K1 (Pi), K2 corresponding to the i th patent document Pi. (Pi), ,,,, Ki (Pi), Kj (Pi) ,,,, Kn (Pi)} may be stored in the core keyword DB 1300 based on Pi or a key value corresponding to Pi. Can be.

Meanwhile, the core keyword meta information generation unit 2140 of the present invention generates at least one or more preset metadata for each of the core keywords, and the generated core keyword meta information is stored in the core keyword meta information DB. The key keyword meta data generated by the key keyword meta information generation unit 2140 is largely two types.

The first is association information between key keywords. All the key keywords are to be associated with the document Pi from which the key keywords appear. Therefore, when there is a patent group generated by at least two patents (an example of the patent group may be composed of all patents, all patents for a certain period, patents by specific patent classification, patents by specific rights holder, patents by specific inventors, etc.). If the key keyword set KS (Pi) corresponding to the patent document Pi is processed by the association group (association analysis, for example, so-called shopping cart analysis), the association information between the key keywords is generated. Association information between keywords can be visualized into the core keyword network. The core keyword network may be analyzed by the network analyzer 4700 of the present invention, and the visualization may be performed by the data visualizer 4700 of the present invention.

The second is the corresponding bibliographic information for each key keyword. The core keyword set KS (Pi) corresponding to the patent document Pi is all of the bibliographic elements of the Pi (application date, applicant, inventor, patent classification, reference, etc.) or by the dispute prediction elements of Tables 1 to 4 corresponding to Pi. The dispute prediction element values are matched. All analysis information will be matched. Therefore, when there is a patent group generated by at least two or more patents (an example of the patent group may be composed of all patents, all patents for a certain period, patents by specific patent classification, patents by specific rights holders, patents by specific inventors, and the like. ) Based on the patent group, at least one or more patent documents can be corresponded to each key keyword. Accordingly, quantitative analysis information (applications, registrations, application increase, registration increase, other patent sets) for the patent documents of the patent group for which the core keywords are the core keywords of the respective core keywords. Quantitative analysis information) may be metadata of the individual key keywords. It is possible to find a rising key keyword and the like using the meta data for each of the key keywords, and by integrating and utilizing the key keyword network, a key keyword network centered on the rising key keyword can be found. On the other hand, since two core keywords correspond to an edge between the core keywords constituting the core keyword network, a patent document including the two core keywords can be searched at the same time. Core keywords can be included in the search index so that only key keywords can be searched separately.)

When individual keywords constituting the core keyword set KS (Pi) corresponding to Pi are processed by association analysis (eg, so-called shopping cart analysis) in units of the patent group, association information between key keywords is generated. Association information between keywords can be visualized into the core keyword network.

Meanwhile, the method of generating a similar patent group using the patent classification by the similar technology patent group generation unit 5123 of the target patent set generation unit 5120 will be described. If the own patent SSi has n IPCs and m unique patent classifications (to be described using USPC as an example), a similar patent group is generated using the n IPCs and the m unique patent classifications. The similar technology patent group generating unit 5123 extracts TS (IPC1) and TS (IPCn) having this IPC, including the main patent classification or the subpatent classification by IPC1 to IPCn. When the size of the extracted TS (IPCi) (the number of patents belonging to TS (IPC1) becomes size) is s (TS (IPC1)), 1 / for TSi (IPCi) belonging to TS (IPCi) The weight of s (TS (IPCi)) is assigned. On the other hand, if the same patent among TS (IPC1) and TS (IPCn) overlaps, that is, if there are two or more IPCs in SSi, and there is a patent in which the two or more IPCs match among TSi belonging to TS, 1 / s (TS (IPCi and IPCj)) may be allocated to the IPCi and at least one IPCj. TS (IPCi and IPCj) refers to TS generated by TSi having both IPCi and IPCj. On the other hand, even if the IPC is not the same IPC in the technical classification scheme may have a similar level or more than a certain level. The similar technology patent group generation unit 5123 extracts at least one or more upper patent classifications, such as upper-level patent classifications of the IPCi of SSi, with reference to the patent classification system, and for the patent documents including the upper patent classifications. Weights that are inversely proportional to the size of the TS including the patent classification may be mapped. Meanwhile, the similar technology patent group generation unit 5123 extracts at least one lower patent classification of the IPCi of SSi with reference to the patent classification system, and the lower patent classification for the patent document including the extracted lower patent classification. The weight may be inversely proportional to the size of the TS that includes a.

On the other hand, the more the application date between the SSi and TSi is more likely to be similar in content even if they have the same patent classification. Therefore, the weight may be multiplied by the obsolescence weight in consideration of the obsolescence function having an inverse relationship to the difference in the filing date. The obsolescence weight may be a constant / f (the number of days elapsed between the filing date of SSi and the filing date of TSi).

On the other hand, when generating similar patent groups using sub-patent classifications other than the main patent classification, when SSi or TSi includes a plurality of patent classifications, the larger the number of patent classifications, the more the patent classification of SSi and the patent classification of TSi match. Even so, the probability of coincidence is low. Therefore, the dilution rate according to the number of patent classifications needs to be introduced. The dilution rate is multiplied by each weight. Dilution rate is inversely proportional to SSi's patent classification number and inversely proportional to TSi's patent classification number. On the other hand, if the matching patent classification in the main patent classification when the dilution rate is applied, the main patent classification matching weight may be further given. That is, when the main patent classification of SSi and the patent classification of TSi coincide with each other, or when the patent classification of SSi and the main patent classification of TSi coincide with each other, a classification matching weight may be additionally given. In the above manner, a preset weight may be given for each TSi generated by using a patent classification. Proprietary patent classifications such as USPC, FT, FI, and ECLA are treated the same as IPC. When TSi corresponding to SSi is associated with two or more types of patent classifications, the weight of TSi may be summed for each patent classification type.

In this manner, a similar patent group may be generated for each patent belonging to the SS. Similar patents may be assigned to individual patents of similar patents that are created. Accordingly, generating a set of similar patent groups in terms of all SSi patents includes: 1) a method for generating patents having a predetermined similarity score or more among similar patent groups generated for each individual patent belonging to the SSi, and 2) individual belonging to the SS. Create a similar patent group including similarity scores for each patent, add the similar patent groups generated, and add the total score of similarity based on the individual patents among the total similar patent groups combined (the similar patents appearing more than once for two or more similarity scores) Based on the number) can be generated a set of similar patent group of a predetermined number or ratio (3 to 5 times the number of patents of the SS, etc.). Through this process, the similarity total scores correspond to each patent included in the similar patent group set. Since the generation of the similarity score also becomes a weight for a kind of TSi, the setting of the similarity score is performed by the weight adjusting unit 5203 of the present invention.

In the set of similar patent groups, a previous set of similar patent groups may be generated. Since there may be a range of patent filing dates when there are a plurality of SSi, a problem arises that it is difficult to uniquely specify the filing date of the SS, which makes it difficult to specify a prior patent group and a similar patent group. A problem arises. The method of generating a set of similar patent groups of the first application includes: 1) a method of generating a similar patent group with similar patents that are earlier than the filing date of the individual patents when generating a similar patent group from individual patents belonging to the SS; Based on the median of the applicant's median of the patents belonging to the patent group, the group of patents having such an attribute is regarded as a pre-applied similar patent group, and the patent group having an attribute as a post-applied pseudo-patent group. There may be a way to do it.

It will be apparent from the above description that the post-appearance similar patent group set may be generated similarly to the method of generating the prior application similar patent group set.

Another representative example of the TS is a user configuration patent set configured by a user. It is possible to limit the term of the preceding patent set or the similar patent group set, the patent set of the first patent application or the post-patent application, and the like by term, attribute (property attribute, patent attribute, technical field attribute, etc.), search term or search expression, and the like. As a general patent set. Thus, the user can create his own patent set by adding his or her own search formula, patent set managed by him, or any patent set limiting means such as his competitor's patent or related company's patent. For example, when there is an owner of A who has attacked himself, a prior patent cited set of specified patent sets, such as a patent belonging to A's patent portfolio or a patent belonging to an attacking patent or a similar patent group of an attacking patent. May be an example of a user configured patent set. Another method of generating a user-configured patent set is to generate a selected method from among the cited patent set and / or similar patent group set and / or patents belonging to a patent set inputted or generated by a user.

On the other hand, when constructing a user-configured patent set, the user may set a subjective weight for the TSi. This weight is referred to as a user set TSi weight. The setting of the weight is performed through a UI provided by the weight adjusting unit 5153 of the present invention. The UI provides a list of patents belonging to the TS, and the user can input a weight, importance, or important grade to the patent list.

Another representative example of the TS is a system configuration patent set provided by the system. The system is a set of negotiated prior-cited patents consisting of pre-patent patents cited by SSi according to attributes representing SS or SSi, such as the most frequent patent classification to which SSi belongs. On the other hand, the system is selected by presenting to the user various patent sets that are set by the administrator of the system, such as patent sets of recent applications or registrations, or patents of universities or research institutes, patent sets of specific rights holders (to be sold, etc.) By default, the system configuration patent set can be generated.

Next, when the SS and the TS are specified, the information processing method of the system will be described. First, the dispute prediction method will be described.

When the TS is determined by the target patent set generation unit 5120, the dispute prediction model value acquisition unit 5130 obtains Sg (TSi), which is the dispute prediction model value for the dispute prediction model Sg, for all TSi belonging to the TS. Save it. The dispute prediction information generating unit 5140 may use Sg (TSi) values for all TSi belonging to TS, but may process only Sg (TSi) belonging to a predetermined high quartile among n quartiles, etc. For example, the Sg (TSi) value may be limited to a processing target such that only those having a predetermined reference value or more are processed. The limitation of the processing target may be set by the system or set by the user.

The dispute prediction information generation unit 5140 adds a predetermined weight to Sg (TSi) through the multiple relationship processing module 5251 with respect to Sg (TSi) obtained by the dispute prediction model value acquisition unit 5130. do. The weight imposed on Sg (TSi) has the same origin as the weight imposed on TSi. The weight imposed on TSi is determined using one or more of 1) relevance weight with SSi, 2) overlapping weight, 3) subjective weight assigned by SSi to user, and 4) subjective weight assigned by user to TSi. The weight imposed on TSi is called W (TSi). The weight imposed on the TSi is also affected by the relationship between the TSi and the SSi. Therefore, W (TSi) may be W (TSi) = W (TSi, SSi). W (TSi, SSi) includes the subjective weights that the user imposes on SSi, and the degree of relevance for each TSi related to SSi (when at least two or more SSi are associated with one TSi (i.e., the same TSi based on SSi) ), A weight relation that takes into account the subjective weight imposed by the user on TSi.

Processing of the information related to the weight is performed through the UI provided by the weight adjusting unit 5503 of the present invention. The user may assign subjective weights to individual patents SSi or TSi through the weight adjusting unit 5503, and may assign weights set by the user according to the type of relevance or the detailed relevance or adjust the weights assigned by the system. For each weighting type of 1) to 4), whether each weighting type is applied or the relative weighting for each weighting type is set (for example, 50% in 1), 30% in 2), and 10% in 3). , 4), 10%, etc.), or the weight assigned by the system to 1) to 4).

The dispute prediction information value generation module 5514 processes the Sg (TSi) and the W (TSi) to generate dispute prediction information wf (Sg (TSi), W (TSi)) for each TSi. Here, wf is a processing formula, and a simple example of wf is Sg (TSi) * W (TSi), and the simplest example may be Sg (TSi) without any weight applied thereto. As such, when the target patent set is determined, the dispute prediction information generation unit 5140 generates wf (Sg (TSi), W (TSi)) for TSi of the target patent set. The dispute prediction information value provider 5150 provides the wf (Sg (TSi), W (TSi)) to the requester. Wf (Sg (TSi), W (TSi)) generated for each TSi is an example of dispute prediction information of the SS position based on the TS. Subsequently, the dispute prediction information generation unit 5140 may generate at least one collective dispute prediction information for the entire target patent set. The collective dispute prediction information generating function is called f (TS). An example of f (TS) is the sum of wf (Sg (TSi), W (TSi)) by TSi, the average of wf (Sg (TSi), W (TSi)) by TSi, the preset quantile or the preset upper percent. Pre-set statistical processing values for the sum or average of wf (Sg (TSi), W (TSi)) per TSi, the distribution of wf (Sg (TSi), W (TSi)) per TSi, and the statistical processing value. An example of f (TS) may be a rating value applied to the set rating model.

38 is a schematic diagram of a process in which the dispute prediction information value generation module 5514 generates a dispute prediction information value for one patent. In FIG. 38, SSi has n TSi and a weight of W (SSi, TSi), respectively. In this case, the dispute prediction information value generation module 5514 may generate dispute prediction information of the following formula for the individual patent SSi.

wf (Sg (TSi), W (TSi)) (SSi) =

{Sg (TS1) * W (SSi, TS1) ++ Sg (TSi) * W (SSi, TSi) ++ Sg (TSn) * W (SSi, TSn)} / {W (SSi, TS1) ++ W (SSi, TSi) ++ W (SSi, TSn)}

On the other hand, the dispute prediction information may generate dispute prediction information such as the following formula for the entire SS consisting of n individual patent SSi.

wf (Sg (TSi), W (TSi)) (SS) = 1- {1-wf (Sg (TSi), W (TSi)) (SS1)} {1-wf (Sg (TSi), W (TSi )) (SSi)} {1-wf (Sg (TSi), W (TSi)) (SSn)}

41 is a diagram illustrating the type and order of dispute prediction information generated by the dispute prediction information generation unit 5140 of the present invention. The dispute prediction information generation unit 5140 may first generate dispute prediction information for each self patent (SL71). That is, the target patent set for one SSi may be generated and then individual TSi or TS-specific dispute prediction information constituting the TS may be generated as the dispute prediction information for the SSi. On the other hand, the dispute prediction information generation unit 5140 may generate the dispute prediction information predetermined for each of a set of self-patents of which at least one is composed of own patents (SL72). Meanwhile, the dispute prediction information generation unit 5140 may generate dispute prediction information preset for each target patent constituting the TS configured for each SSi or SS (SL73). In addition, the dispute prediction information generation unit 5140 may generate dispute prediction information preset for each target patent set configured for individual SSi or SS (SL74).

Next, the dispute prediction information generation method to which the aggregation concept is applied will be described with reference to FIGS. 42 to 43. 42 illustrates an exemplary method of dividing a self-patent set through the patent set divider 5310 of the present invention and then generating dispute prediction information for the divided self-patent set. The dispute prediction engine 5100 obtains a self patent set (SL81), obtains a division criterion for the self patent set (SL82) through the patent set divider 5310, and corresponds to the divided self patent set. After generating the target patent set SL83, dispute prediction information based on the target patent set corresponding to the divided self patent set is generated (SL84).

FIG. 43 illustrates an exemplary method of dividing a target patent set through the patent set dividing unit 5310 of the present invention and then generating dispute prediction information for the divided target patent set. The dispute prediction engine 5100 obtains a self patent set (SL91), generates a target patent set corresponding to the self patent set (SL92), and generates a dispute prediction model value for each target patent (SL93). Subsequently, through the patent set dividing unit 5310, after obtaining a dividing criterion, the target patent set is divided (SL94), and dispute prediction information generation (SL95) is generated based on the divided target patent set. do. Hereinafter, powdering is performed.

The TS is a target patent set, and the target patent set may be divided into at least one unit by applying at least one or more division criteria. The division may be performed by the patent set division unit 5310 of the dispute prediction information analysis engine 5300 of the present invention. Let the i-th TS divided into units be TSui. In this case, the dispute prediction information generation unit 5140 may generate a value of wf (Sg (TSui), W (TSui)) or f (TSui) for each TSui. How to segment TS is 1) by owner, 2) by technology group (may use IPC or USPC), 3) by period, 4) by inventor, 5) by country, 6) by owner attribute, 7) at least By combination of one or more specific keywords, 8) by conditions that can divide or define other patent sets, 9) by patent groups with specified attributes, and / or 10) by combining the methods of 1) to 9) above. There may be. Examples of 6) may be (1) registered troll attributes, (2) multidispute triggers, and the like. Examples of the specified attribute of 9) include (1) a patent group having a high similarity to SS among TS, and (2) a patent consisting of TSi corresponding to SSi whose user-set SSi weight is higher than a predetermined ranking, level, and range. Military and the like.

Meanwhile, the SS is a self-patent set, and the self-patent set may be divided into one or more units through a division method as the target patent set. The divided i-th SS may be called SSui, and a TS (SSui) which is a target patent set generated for each SSui may be generated, and wf (Sg (TS (SSui)) for each TS (SSui) with respect to the TS (SSui). )), W (TS (SSui))) or f (TS (SSui)). On the other hand, the TS (SSui) is also a target patent set is a kind, can be divided by the partitioning method, and also for the divided TSui (SSui) wf (Sg (TSui (SSui)), W (TSui (SSui)) )) Or f (TSui (SSui)).

The dispute prediction information analysis engine 5300 of the present invention analyzes the dispute prediction information. FIG. 44 shows an exemplary configuration of an information processing method of the dispute prediction information analysis engine 5300 of the present invention. The dispute prediction information analysis engine 5300 generates the dispute prediction model value information for each target patent (SL101), and then target patents, target patent sets, and split target patents for respective patents, own patent sets, and divided self patent sets. The dispute prediction model value or the dispute prediction information value based on the set is generated (SL102). Subsequently, the dispute prediction information analysis engine 5300 performs quantitative analysis of ranking, distribution, right holder, right holder type, inventor, and patent classification of each dispute prediction model value or dispute prediction information value for each criterion (SL103), or a target. Quantitative analysis is performed for each bibliographic item or analytical index of the patent (SL104) to generate a quantitative analysis result.

 The dispute prediction information to be analyzed by the dispute prediction information analysis engine 5300 is SSi, TSi, wf (Sg (TSi), W (TSi)), f (TS). In particular, TSi has a bibliographic value and a dispute prediction element value for each dispute prediction element of Tables 1 to 4, wherein the bibliographic value, dispute prediction element value, and wf (Sg (TSi), W (TSi)) are various. May be subject to quantitative analysis. For example, with the wf (Sg (TSi), W (TSi)) value for TSi for each rights holder included in TS, it is possible to analyze how much dispute prediction information value a specific rights holder has in SS, A ranking of wf (Sg (TSi), W (TSi)) may be generated, such as generating TSi having the largest wf (Sg (TSi), W (TSi)) value with respect to TSi. This effectively identifies patents and rights holders who are likely to attack companies that own the SS or produce products related to the SS. On the other hand, by analyzing the TSi having a predetermined wf (Sg (TSi), W (TSi)) or more to find the least inventor, it is also possible to generate wf (Sg (TSi), W (TSi)) information for each of the least inventors . On the other hand, the TSi of the individual owner or SME may be extracted from TSi whose wf (Sg (TSi), W (TSi)) is greater than or equal to a predetermined standard.

When the dispute prediction function value or the dispute prediction function value is generated for each f (TS) or f (TSui) for f, which is a dispute prediction function, how to structure the dispute prediction function value to generate dispute prediction information is the dispute prediction information of the present invention. The dispute prediction model, which is performed by the generation unit 5140, may be used by the dispute prediction model. Meanwhile, how to output the dispute prediction result is performed by the dispute prediction result output unit. The actual report in pdf form or e-mail report is generated by the patent dispute information report generator 4440 of the present invention.

Next, a method of generating, by the dispute prediction engine 5100, predetermined grade information based on the dispute prediction information value will be described with reference to FIG. 45. The dispute prediction engine 5100 generates a dispute prediction model value or a dispute prediction information value based on a target patent, a target patent set, and a split target patent set for each of a self patent, a self patent set, and a divided self patent set (SL111). ), And generates grade information according to the graded model (SL112). If a dispute prediction information value or a dispute prediction function value is generated for each of f (TS) or f (TSui) with respect to f, which is at least one collective dispute prediction information generating function, the dispute prediction grade is determined according to a preset rating model. Can be. The grading model may assign a rating to each dispute prediction function value section, or determine a rating section based on the distribution in consideration of the distribution of the dispute prediction function value, and give a prediction rating to the determined rating section. When there are score values, there are a number of models that divide them into n classes, which will be apparent to those skilled in the art.

Next, a method of upgrading the model of the patent dispute prediction information generation system 5000 will be described. If dispute-prone patents are added over time, a dispute prediction model can be generated from the existing dispute prediction element or a new dispute prediction element set by including the additional dispute occurrence patents, and the dispute prediction information can be generated using the generated dispute prediction model. Will be. In other words, the dispute prediction element value for each Pi changes over time, and if the dispute prediction element value is different, the dispute prediction model is different, and the dispute prediction model value and the dispute prediction information value are also different. The patent dispute prediction information generation system 5000 may predict at least one or more disputes regarding at least one patent set managed by the patent dispute prediction information generation system 5000 and at least one patent set managed by the user. It is possible to batch generate the model value and the dispute prediction information value according to a predetermined cycle or a predetermined condition.

Next, with reference to FIGS. 46-47, the information processing method of the dispute propagation prediction engine 5600 of this invention is demonstrated. The dispute propagation prediction engine 5600 obtains a new individual dispute patent or a new dispute patent set (SL121), and then generates a target patent set for each individual dispute patent obtained or for each new dispute patent set (SL122). . Subsequently, the dispute propagation prediction engine 5600 performs a predetermined quantitative analysis including an owner analysis on the target patent set (SL123). The dispute propagation prediction engine 5600 obtains and registers (SL131) a condition for generating at least one self-patent set or self-patent set from a user, and then sets a self-patent set or a self-patent when a periodic or preset condition is satisfied. Dispute prediction information is generated for each patent of the patent set (SL132). Subsequently, the dispute propagation prediction engine 5600 reports the generated dispute prediction information to the user (SL133). Hereinafter, powdering is performed.

When at least one new dispute patent is obtained, the dispute spread prediction engine 5600 analyzes the dispute patent through a new dispute analysis unit 5610 to generate analysis information regarding the new dispute patent, or Alternatively, the newly generated dispute propagation prediction unit 5620 may generate dispute propagation prediction information about the new dispute patent, or may perform notification processing on the new dispute patent through the dispute propagation notification unit 5630.

First, the information analyzed by the newly generated dispute analysis unit 5610 includes a time series distribution of the number and the number of times the new dispute patent caused a dispute, whether a dispute party (defendant, etc.) overlaps, and the new dispute patent is included. Analyze the time series distribution of disputes between parties in disputes. On the other hand, a predetermined quantitative analysis process for a new disputed patent on a predetermined period basis. The quantitative analysis process includes generating quantitative analysis information by right holder, dispute claimant, dispute patent, defendant, inventor, patent classification, and dispute filing date.

The new occurrence dispute spreading unit 5620 treats each new disputed patent as a self patent (SSi), and uses the target patent set generation unit 5120 of the present invention to apply for the own patent application (the best date). Or a priority date.) A TS composed of a post-application patent is generated as a reference. A predetermined quantitative analysis process is performed on the generated TS, including an owner analysis of TSi. On the other hand, the new occurrence dispute spreading unit 5620 constitutes a self-patent set of a new dispute patent set consisting of two or more new disputed patents, and the target patent set generation unit 5120 for the self-patent set. By using, a TSi (SSi) which is a post-application is generated on the basis of the filing date of the SSi of the self-patent set (which may be the best date or the priority date), and the TS is composed of the generated TSi (SSi). A predetermined quantitative analysis process is performed on the generated TS, including an owner analysis of TSi. At this time, the TSi constituting the target patent set is given that the predetermined dispute prediction model value and weight of the present invention are given, and the dispute prediction model value and the weight are used for the quantitative analysis of the present invention. It will be natural.

Next, a method of providing an alert service for each user using the fraud dispute spreading notification unit 5630 will be described. The user may set at least one SS, and may register rights holders, such as competitors associated with themselves, own interest patents, own interest technology groups (registered in a patent classification), and the like. In this case, when a change occurs in the entire patent data such as the generation of new patent data or a change in the owner, or when a new troll is added, a standard patent pool is added, or a dispute-prone patent is added periodically, The dispute prediction function value is changed for each of f (TS) or f (TSui). At this time, the dispute prediction function for each f (TS) or f (TSui) at a specific time point (t = t1) and the dispute prediction function for each f (TS) or f (TSui) at a different time point (t = t2) If the value is greater than or equal to the preset value, the system can provide information about the occurrence of the change to the user. For example, in the last month, when a troll has an event where a troll is a subsidiary of the troll and has purchased n patents from at least one other person (which can be identified by current assignee information), More than a predetermined value is changed in the dispute prediction function value for each f (TS) or f (TSui) for SS at the present time (t = t2), so that the risk of patent disputes may increase significantly for the user. . Of course, the system may perform a predetermined quantitative analysis (author analysis, technical field analysis, time series analysis, etc.) for the patent when there are n patents, and parent patents cited by the n patents A set can be obtained, and a predetermined quantitative analysis can be performed for the parent patent set.

Next, with reference to FIG. 48, the risk hedging prediction information generation is demonstrated using the patent dispute prediction information generation system 5000 of this invention. The risk hedging information generation unit of the present invention is responsible for generating the risk hedging prediction information. 48 is a diagram showing the configuration of an information processing method of the risk hedging information generation unit. The risk hedging information generation unit first obtains a target patent set including at least one patent and obtains at least one complementary patent set having a predetermined relationship with the target set, or first obtains a complementary patent set including at least one patent. And acquire at least one or more target patent sets having a predetermined relation with the complementary set (SL141), obtain at least one dispute prediction model value for each patent for each individual patent constituting the supplementary patent set (SL142), and for each patent. At least one risk hedging prediction information is generated using the dispute prediction model value (SL143). It will be described in more detail below.

In the risk hedging prediction information, when there is a TS related to the SS, a risk hedging set (RHS) is a kind of complementary patent set that can predict a patent attack on the whole TS or at least one divided TS. ) And cross-licensing set (CLS), which is a kind of complementary patent set having a predetermined relationship designated by SS or SS generator among them. First, the method of generating the RHS by the risk hedging information generation unit will be described. The risk hedging information generation unit obtains selection information of the user or the patent dispute prediction information generation system 5000 for the whole TS or the divided TS of at least one director. The divided TS may be divided TSs according to the right holders or the right holders having a specific attribute (troll, competitor, dispute claimant, dispute experienced person, etc.) among the TSs, and among them, the wf (Sg ( TSi) and W (TSi)) may be divided TSs containing a large number of divided TSi, or divided TSs having a high f (TS) for the divided TSs.

The risk hedging information generation unit may treat the TS or the divided TS (hereinafter referred to as re-insert TS) as SS. When the re-entry TS is treated like a self-patent set, the target patent set generating unit 5120 generates a TS (re-entry TS) which is a target patent set for the re-entry TS. The dispute prediction element value generator 5510 generates a dispute prediction element value for each dispute prediction element of Tables 1 to 4 with respect to TSi (re-entry TS), which is a target patent constituting the TS (re-entry TS). do. The dispute prediction model value acquisition unit 5130 obtains a dispute prediction model value for the generated dispute prediction element value of the TSi (re-entry TS). The dispute prediction information generation unit 5140 generates all predetermined information about the TSi (re-entry TS). The information generated by the dispute prediction information generator 5140 may be wf (Sg (TSi (re-input TS)), W (TSi (re-intake TS))) or f (TS (re-intake TS)).

The risk hedging information generation unit has a patent in which wf (Sg (TSi (re-input TS)), W (TSi (re-intake TS))) of TSi (re-input TS) is equal to or higher than a predetermined level, or wf (Sg (TSi). Complementary patents that are patents of right holders whose (re-entry TS)), W (TSi (re-entry TS))) are above the predetermined level and the applicant becomes an individual, SME, university, institution, or has a predetermined risk hedging attribute Candidate or complementary patent groups are selected. The risk hedging attribute is an attribute that can hedge risk from the point of view of the SS or the creator of the SS. The risk hedging attribute refers to an attribute capable of countering a potential attacker or reducing the risk of other patent disputes. It becomes a strong complementary patent. If a TS composed of patents of a specific right holder with a high risk of patent disputes in the SS has a high risk of dispute from a particular patent group A, the particular patent group A is a supplement that is likely to hedge the risk from the TS for the SS. It becomes a patent group. Therefore, finding the supplemental patent group A as described above is information processing performed by the risk hedging information generating unit.

The division TS is performed by the patent set division unit 5310. If there is a divided TS of the specific rights holder, the risk hedging information generation unit controls the patent set dividing unit 5310 to generate at least one redivided TS according to various division policies for the divided TS of the specific rights holder. The risk hedging information generation unit for each of the repartitioned TSs includes a patent in which wf (Sg (TSi (repartitioned TS)) and W (TSi (repartitioned TS))) of TSi (repartitioned TS) becomes a predetermined level or higher, or wf (Sg ( As TSi (Redivision TS), W (TSi (Redivision TS)) is above the predetermined level, the applicant becomes an individual, a small business, a university, an institution, or selects a patent of a right holder having a predetermined risk hedging attribute. It may be a kind of simulation that the risk hedging information generation unit generates a plurality of repartitioned TSs and discovers the patent group A for the generated repartitioned TSs. On the other hand, the redistribution may apply clustering to the divided TSi constituting the divided TS of the specific rights holder, to configure n clustered redistributed TSs, or to divide the divided TSi by patent classification to generate a plurality of redistributed TSs. will be. The information processing by the risk hedging information generation unit for each of the repartitioned TS may find counterattackable patent group A targeting a wide range of the divided TS, in consideration of the counterattack against the owner of the divided TS. This is because they can find a specialized but powerful counterattack group B that targets a narrow range.

Next, with reference to FIG. 49, the information processing method of the cross licensing information generation part 5340 of this invention is demonstrated. The cross-licensing information generation unit 5340 first obtains a target patent set including at least one or more patents, and obtains at least one complementary patent set having a predetermined relationship with the target set, or includes a complementary patent including at least one patent A set is first obtained, and at least one target patent set having a predetermined relationship with the complementary set is obtained (SL151), and then, at least one dispute prediction model value for each patent is obtained for the individual patents constituting the supplementary patent set (SL152). In operation S153, at least one cross-licensing prediction information is generated by using a dispute prediction model value for each patent. It will be described in more detail below.

The cross licensing patent group is generated by the cross licensing information generation unit 5340 of the present invention. The cross-licensing information generation unit 5340 is included in the SS from the patent group A that can counterattack a wide range of the divided TS or the patent group B that is capable of counterattacking a narrow range of several subdivision TSs. Or extract the patent of the right holder (the most representative one can be company, subsidiary, subcontractor, original government or subcontractor, individual, institution, university, etc.) with special relationship. On the other hand, the cross-licensing information generation unit 5340 may be assigned to the divided TS, the SS (or the subsidiary, subcontractors, etc.) designated by the person generating the SS when the TS (divisional TS) is generated And the patent of the original government office, a partner company, an individual, an institution, a university, etc. having a special relationship. Only a patent having a desired attribute can be applied to the target patent set generation unit 5120. (Split TS) can be generated. On the other hand, the cross-licensing information generation unit 5340 controls the dispute prediction information generation unit 5140 with respect to wf (Sg (TSi (divisional division)) for TSi (divisional TS) which is an individual patent constituting the TS (divisional TS). TS)), W (TSi (split TS)) or f (TS (split TS)). In general, there is a high possibility that there is no patent matching the SS and the TS (split TS), but there are many patents constituting the SS, and a wide range of filing dates of the patents constituting the SS are not excluded. can not do it.

Next, a description will be given of a configuration of an application system utilizing each engine, a DB, and a functional unit constituting the patent dispute prediction information generation system 5000 of the present invention, and an information processing method of the application system. A representative application system is a patent licensing prediction model generation system. 50 illustrates an exemplary configuration of a patent licensing prediction information generating system 6000, and FIG. 51 illustrates an exemplary configuration of an information processing method of the patent licensing prediction information generating system 6000.

First, the patent licensing prediction information generating system 6000 will be described. When a patent dispute occurs, the dispute is often terminated by licensing, and when a licensing negotiation fails, it is often turned into a dispute. Thus, patent disputes are a two-sided relationship between licensing and coins. Therefore, if most of the functions of the patent dispute prediction information generation system 5000 are utilized as it is, the patent licensing prediction information generation system 6000 can be configured, and most of the functional modules of the patent licensing prediction information generation system 6000 are used. The same works in the case of the patent dispute prediction information generation system 5000. However, in the case of the patent licensing prediction information generation system 6000, unlike the patent dispute prediction information generation system 5000, when the target patent set generation unit 5120 generates a similar patent group having a predetermined relationship, There is a big difference that a target patent set is generated for TSi, which is a later application than SSi of a patent set.

 In the relationship between the patent dispute prediction information generation system 5000 and the patent licensing prediction information generation system 6000, the dispute prediction element may correspond to the licensing prediction element, and the dispute prediction model may correspond to the licensing prediction model, and the dispute DB unit ( 5200 is a licensing DB unit 6200, a dispute prediction information analysis engine 5300 is a licensing prediction information analysis engine 6300, a dispute prediction management unit 5400 is a licensing prediction management unit 6400, a dispute prediction model generation engine 5500 Corresponds to the licensing prediction model generation engine 6500, the dispute prediction engine 5100 is a licensing prediction engine 6100, the dispute prediction model value acquisition unit 5130 is a licensing prediction model value acquisition unit 6130, dispute prediction The information generation unit 5140 is a licensing prediction information generation unit 6140, the dispute prediction information value generation module 5152 is a licensing prediction information value generation module 6162, and the dispute prediction information value providing unit 5150 is licensing prediction information. The value providing unit 6150, the dispute prediction element value DB 5220, the licensing prediction element value DB 6220, the dispute prediction model value DB 5230, the licensing prediction model value DB 6230, and the dispute prediction system manager 5520. The licensing prediction system manager 6620, the dispute prediction information batch generator 5241, the licensing prediction information batch generator 6421, and the dispute prediction user manager 5430 are the licensing prediction user manager 6630, and the dispute UI unit 5431. ) Is a licensing UI unit 6431, the dispute prediction element value generator 5510 is a licensing prediction element value generator 6510, and the dispute prediction model generator 5520 is a licensing prediction model generator 6520, by patent. The dispute prediction model value generating unit 5530 corresponds to a licensing prediction model value generation unit for each patent, the dispute prediction model value providing unit 5540 corresponds to the licensing prediction model value providing unit 6540, and the attack prediction information generation unit corresponds to the licensing prediction information generation unit. Patent licensing prediction information Description of the configuration and the information processing method of the system (6000) is sufficient to one of ordinary skill in the art that changes in the description of the patent conflict prediction information generating system (5000) with the corresponding terms are described above.

Meanwhile, in the patent licensing prediction information generation system 6000, the dispute occurrence patent DB 5210 and 6210 and the dispute data acquisition unit 5410 and 6410 of the patent dispute prediction information generation system 5000 are used in the same manner. Patent set generation unit 5110 and 6110, target patent set generation unit 5120 and 6120, cited patent set generation unit 5121 and 6121, similar patent group generation unit 5122 and 6122, and similar technology patents Group generation unit 5123 (6123), target set acquisition unit (5124) (6124), option processing unit (6125) (5125), multiple relationship processing module (5141) (6141), weight adjusting unit (5143) (6143) ) Is used equally. In particular, when the patent licensing prediction information generation system 6000 generates a licensing prediction element and a licensing prediction model, the patent licensing prediction information generation system 6000 generates a dispute using a dispute-prone patent and a non-dispute-prone patent, and thus disputes with the dispute-prone patent DB 5210 and 6210. The data acquisition units 5410 and 6410 are maintained under the same names as the patent dispute prediction information generation system 5000. In this paragraph, the same reference numerals are used for the components used in the patent licensing prediction information generating system 6000 and the patent dispute prediction information generating system 5000 without changing their names.

As shown in FIG. 51, the patent licensing prediction information generating system 6000 first obtains a self-patent set including at least one or more patents, and obtains at least one or more target patent sets having a predetermined association with the self-set, First, obtain a target patent set including at least one patent, obtain at least one self patent set having a predetermined relationship with the target set (SL161), and at least one or more patents for individual patents constituting the target patent set. Each license prediction model value is obtained (SL16), and at least one license prediction information is generated (SL163) using the license prediction model value for each patent. The generation of the licensing prediction information of the patent licensing prediction information generating system 6000 applies the method of generating dispute prediction information of the patent dispute prediction information generating system 5000, and the individual licensing prediction information generating method applied mutatis mutandis is described above. Since it will be apparent to those skilled in the art through the above-described method of generating dispute prediction information, a detailed description thereof will be omitted.

Next, the method of generating the group unit citation analysis information which is one of the element technologies which are the core of the patent dispute information processing will be described in detail with reference to the drawings.

6 illustrates one exemplary configuration for an aggregated unit citation analysis 4500 of the present invention. The aggregated unit citation analysis 4500 acquires a patent set composed of at least two or more pieces of patent information and generates an acquired patent set, and an individual patent included in the obtained self-patent set. A target patent set generation unit 4520 for generating at least one cited patent set by processing a citation analysis unit 4530 for processing a citation analysis based on at least one cited analysis purpose set for the cited patent set; It includes.

The acquisition patent set generation unit 4510 includes a patent set acquisition unit 4511 for acquiring a patent set and an acquisition patent set limitation unit 4512 for limiting the acquired patent sets. The acquisition target patent set generation unit 4520 includes a target patent set acquisition unit 4522 for acquiring the target patent set and a target patent set limitation unit 4451 for defining the target patent set. The patent set obtained by the target patent set obtaining unit 4522 may include a forward cited patent set, a back cited patent set, a front self cited patent set, a back self cited patent set, a citation generation obtained patent set, and the like. When the acquisition target patent set generation unit 4520 defines a target patent set by applying specific conditions, the target patent is obtained from the data unit 1000 by reflecting the above conditions. The cited patent set generation unit 4523 generates a cited patent set as the obtained target patent group. The citation analysis unit 4530 includes a citation analysis purpose selection unit 4531 for selecting a citation analysis purpose and a citation analysis execution unit 4452 for performing citation analysis for the selected purpose. Meanwhile, the patent information system 10000 may generate and manage patent sets for at least one or more categories, and the patent set management unit of the present invention manages the patent sets. At the bottom of the system patent set management unit, an applicant-specific patent set management unit managing a patent set specific to each applicant, a patent set management unit by patent classification managing a patent set specified by patent classification, and a patent set specified by other classifications or categories There are other category-specific patent set management departments to manage.

On the other hand, the patent set generated by the user is performed by the member patent set management unit.

As illustrated in FIG. 2, the patent data unit 1100 includes a patent specification file unit 1110, a patent DB 1120, a patent classification DB 1130, and the like. The patent DB 1120 manages bibliographic matters of all patents, the specification body, and drawings for each field, and various fields constituting the specification body (title, summary, prior art, claims, detailed description of the invention, etc.). Contains key keywords extracted from. Meanwhile, the patents may further include citation information as prior art documents for the patents. For example, in the US patent data, the citation information is information included in the reference, and includes a US patent document number, a foreign patent document number, and an indicator of a non-disputed patent document. On the other hand, prior art research information by examiners or related persons of the Patent Office, and information on the examination when submitting the examiner's opinion are broad citation information. If there is forward citation information in a specific document, the specific document is a backward citation document from the viewpoint of the document included in the forward citation information. The document included in the forward citation information based on the specific document becomes the parent document and the specific document becomes the child document based on the parent document. It is obvious to those skilled in the art to process the information in the child-parent relationship with the DB, and a detailed description thereof will be omitted.

The bibliography of the patent document includes the issuing country information, various date information, various number information, at least one owner information, at least one inventor information, at least one patent classification information, at least one priority information, and the like. The date includes the filing date, the date of publication, the date of registration, and other dates. The various number information includes an application number, a public number, a registration number, an original number, and a priority claim number. The right holder information includes the applicant, the assignee, the patent owner, etc. If there is a change in the right holder and the change is managed, there may be information about the transferor and the assignee and information about the right holder. Priority information includes information such as priority claim number, date of claim, and country. On the other hand, when there is a split application, partial continuation application, continuation application, etc., information such as an original number, date of the original source is added. In addition, bibliographies may include representations, titles, abstracts, and index words. Meanwhile, the bibliographies processed include domestic family information (such as a split application, a change of application, or a part of a continuation application, a patent application that is related to a continuation application), overseas family information (an application that may be related to a treaty priority relationship, This can be. On the other hand, the text of the patent specification body may be further extracted by a keyword extracting method for each field or combination of fields constituting the main text through natural language processing or the like. In addition to the common IPC, the patent classification information may have a unique patent classification for each country such as USPC, FT, FI, ECLA, etc.

Next, the information processing method of the aggregated unit citation analysis 4500 will be described in more detail.

The acquisition patent set generation unit 4510 receives an acquisition patent set consisting of at least two or more pieces of patent information, and the target patent set generation unit 4520 processes individual patents included in the acquisition patent document set. Generate at least one target patent set, and the citation analysis unit 4530 processes at least one citation analysis preset for the target patent set.

On the other hand, as can be seen in Figure 20, the set unit citation analysis 4500 is limited to the acquisition patent set or the target patent set through the acquisition patent set limiting unit 4512 or option selection unit 4340, Obtain selection information on the purpose of citation analysis through the citation analysis purpose selection unit 4531, and generate a citation patent set that is the target of citation analysis for a limited patent set through the citation patent set generation unit 4523, The citation analysis purpose selection unit 4531 and the citation analysis execution unit 4532 generate analysis information according to the citation analysis purpose for the patents included in the citation analysis target set.

The citation analysis unit 4530 of the present invention obtains a selection for a citation purpose from a user by using the citation analysis purpose selection unit 4531. The purpose of citation analysis is to indicate what key information appears in the citation analysis results. Examples of citation analysis purposes may include total amount, applicant / owner, inventor, patent classification, and individual patent. When there is a preset patent set related to the citation, the citation analysis execution unit 4532 may include the total patent amount in the patent set, the total amount of patents by applicant or right holder, the total amount of patents by time, the total amount of patents by patent classification or the total amount of patents by time, etc. At least one or more of the same preset quantitative analysis may be performed. The citation analysis execution unit 4532 includes a quantitative analysis unit that performs quantitative analysis. The quantitative analysis may be performed for each field. Meanwhile, the quantitative analysis may include time series analysis, and the time series analysis for each field is also included in the quantitative analysis.

The citation direction can be either forward or backward or the citation generating patent set itself. On the other hand, the forward cited patent set or the back cited patent set is reasonable to allow duplication, but in a special case, the user may select not to allow duplication. That is, when n patents (I1 ,,,,, In) belonging to the acquired patent set refer to the same patent Pi, it is more appropriate that Pi should be treated as having a weight of n times. Typically, when performing a set operation (set operation), overlapping elements (Pi) are treated as one time by eliminating duplication, but for the purpose of citation analysis, this is not valid. Therefore, each Pi must be processed while maintaining frequency n.

On the other hand, rising analysis (increase / decrease rate analysis) for the patent group constituting the forward-cited patent set, the back-cited patent set, the forward self-cited patent set, the back-self-cited patent set, the citation generation obtained patent set, etc., with or without limitation. ) Can be performed. The rising analysis is time series analysis of individual applicants, inventors, patent classifications, keywords (including keyword pairs), individual documents or co-applicants, joint inventors, patent classification pairs, and frequency by keyword pairs. You can do it. On the other hand, new entry analysis is performed on the patent group constituting the above-mentioned or unrestricted forward cited patent set, back cited patent set, forward self cited patent set, back self cited patent set, cited generation obtained patent set, and the like. Can be done. A new entry analysis is defined by a specific cutoff (threshold) at a specific point in time, and includes individual applicants, inventors, patent classifications, keywords, individual documents or co-applicants, co-inventors, patent classification pairs, You will be able to extract keyword pairs. Meanwhile, as a type of new entry analysis, applicants, inventors, patent classifications, keywords, individual documents or co-applicants, co-inventors, patent classification pairs, and applicants, inventors, and patent classifications having high rising speeds of keyword pairs based on specific cutoffs. , Keywords, individual documents or co-applicants, common inventors, patent classification pairs, keyword pairs.

On the other hand, every number in the results of the citation analysis can be assigned a condition that the number appears in (all conditions can be limited to SQO statements). It will be easy for one skilled in the art to be able to call. In this case, the citation analysis can be performed again using the patent document corresponding to the analysis value as the acquired patent set.

Meanwhile, the network analyzer 4700 may analyze a network for an association between the joint defendants, and the data visualization unit 4710 may visualize the network between the joint defendants. For example, for patent dispute i, if defendant i and at least one defendant j are joint defendants or if dispute i and at least one defendant j are joint defendants, at least one patent It is possible to perform a network analysis between joint defendants by a patent dispute group including dispute i or by a dispute patent group including at least one dispute patent i. Meanwhile, since the network analysis unit 4700 has a manuscript for each dispute patent i or a patent dispute i, each network dispute unit 4700 includes a patent dispute group including at least one patent dispute i or a dispute patent group including at least one dispute patent i. Network analysis can be performed, including plaintiffs and joint defendants. At this time, the relationship between the plaintiff and the defendant is directional, and the relationship between the co-defendants can be processed without the directivity. In this case, when the network analysis is performed for each patent dispute group or dispute patent group, the number of the manuscripts is the largest, so the manuscript is at the center. The network analyzer may perform network analysis on any relation. The association analyzed by the network analyzer 4700 may be based on any patent set, when there are a plurality of key keywords included in the same patent document, when there are at least two or more patent classifications, or at least two or more patents. The same applies to the case where there is an inventor, at least two applicants, or at least two cited patent documents. As such, when two or more objects exist in a single object at the same time, when there are n single objects, the two or more objects that exist at the same time have an association relationship, and are subject to network analysis.

The present invention can be widely used in the patent information industry, patent information analysis industry, legal industry, patent consulting industry, R & D and the like.

100: user computer
1000: data section
10000: Patent Information System
1100: Patent data part
1110: Patent specification file portion
1120: Patent DB
1130: Patent Classification DB
1140: Other Patent Data DB
1200: non-patent data portion
1210: non-patent file portion
1220: Non-patent DB
1230: Company Information DB
1240: other non-patent data DB
1300: Key Keyword DB
131-1: Technical Keyword Air Pair DB
131: Technical Keyword DB
1310: Core Keyword Metadata DB
1311: Keyword Relationship DB
1312: representative keyword DB by patent classification
132-1: Product Keyword Air Pair DB
132: Product Keyword DB
133-1: Keyword Syntax Air Pair DB
133-2: Syntax Keyword Air Pair DB
133-3: Syntax DB by Pattern
133: syntax keyword DB
1400: classification metadata DB
1700: user data section
1800: other DB
1900: Purpose-specific data section
1910: Applicant DB
1920: rule data portion
1930: language data section
1940: Standard Patent Pool Data Section
200: wired and wireless network
2000: Data Processing Department
20000: Purpose Specialized System
2032: citation analysis data generation unit
2100: key keyword generator
21000: Nich Technology Zone Discovery System
2110: keyword extraction unit
2120: Key Keyword Selection Unit
2130: syntax keyword generation unit
2140: core keyword meta information generation unit
2150: product keyword generator
2200: classification metadata generation unit
22000: Patent Auto Evaluation System
2210: classification keyword core keyword group generation unit
2220: mapping information generation unit between classifications
2230 key keyword stratification unit
2240: Patent automatic classification unit
2300: purpose-specific data generation unit
23000: Future Technology Prediction System
2310: statistical data generation unit
2330: data generation unit for discovery analysis
2400: Similar patent set generation unit
24000: Convergence Information Processing System
2410: clustering processing unit
2500: network data generator
2510: node data generator
2520: edge data generator
300: linkage system
3000: Patent Information Service Support Department
3100: search processing unit
3110: search engine unit
3120: DB query processing unit
3200: member management unit
3210: Member data
3211: Membership Document DB
3212: Member mapping information DB
3213: Member Billing DB
3220: member directory generator
3230 member flag generator
3240: Member mapping information generation unit
3241: member keyword mapping information generation unit
3242: Member Applicant Mapping Information Generator
3300: platform service provider
3310: Web service provider
3311: Web service account management unit
3312: Web service interface unit
4000: patent analysis information generation unit
4100: acquisition target
4200: Patent Analysis Processing Unit
4210: Analytical Indicators DB
4220: Analysis command syntax DB
4300: analysis option processing unit
4310: data limited option processing unit
4310: Comparative Analysis
4320: what if analysis unit
4320: display option processing unit
4330: other option processing unit
4330: Transfer / Consultation Analysis Department
4340: association analysis
4340: Option selector
4341-1: Limited Time
4341-2: State Limited
4341-3: Applicant Limited
4341-4: Inventor Limited
4341-5: Patent Classification Limited
4341-6: Field Attribute Qualifier
4341-7: Individual field value limiter
4341-8: Individual patent attribute value limiter
4341-9: Other Limited
4350: Risk / Opportunity Analysis Department
4360: 2x2 analysis unit
4370: Open Innovation Target Discovery Department
4400: Analysis result reporting unit
4410: table generator
4420: chart generator
4430: diagram generator
4440: report generation unit
4500: set unit citation analysis unit
4510: acquisition patent set generation unit
4511: patent set acquisition department
4512: acquisition patent set limited part
4520: target patent set generation unit
4521: limited target patent set
4522-1: Acquisition of cited patent set
4522-2: Obtained citation patent set
4522: acquisition of target patent set
4523: cited patent set generation unit
4530: citation analysis unit
4531: Citation Analysis Purpose Selection Unit
4532: Citation Analysis Execution Unit
4600: Analysis result reporting unit
4700: network analysis unit
4710: data visualization
5000: Patent dispute prediction information generation system
5100: Dispute Prediction Engine
5110: self-patent set generation unit
5120: target patent set generation unit
5121: Citation Patent Set Generator
5122: similar patent group generation unit
5123: similar technology patent group generation unit
5124: target set acquisition unit
5125: Option processing unit
5130: obtain a dispute prediction model value
5140: dispute prediction information generation unit
5141: Multiple Relationship Processing Module
5142: dispute prediction information value generation module
5143: weight control unit
5150: dispute prediction information value provider
5200: dispute DB
5210: Dispute Generation Patent DB
5220: dispute prediction element value DB
5230: dispute prediction model value DB
5300: Dispute Prediction Information Analysis Engine
5310: Patent Set Division
5320: attack group information generation unit
5330 risk hedging information generation unit
5340: cross-licensing information generation unit
5400: Dispute Prediction Management
5410: Obtaining Dispute Data
5420: Dispute Prediction System Management Department
5421: Dispute prediction information batch generation unit
5430: Dispute prediction user management
5431: Dispute UI Department
5432: User input patent group management unit
5500: Conflict Prediction Model Generation Engine
5510: dispute prediction element value generation unit
5520: dispute prediction model generator
5530: dispute prediction model value generation unit for each patent
5540: dispute prediction model value provider
5600: Dispute Ripple Prediction Engine
5610: New occurrence dispute analysis
5620: New Outbreak Dispute Prediction Unit
5630: Dispute Ripple Notification Unit
6000: Patent Licensing Prediction Information Generation System
6100: Licensing Prediction Engine
6110: self-patent set generation unit
6120: target patent set generation unit
6121: Citation Patent Set Generator
6122: similar patent group generation unit
6123: similar technology patent group generation unit
6124: target set acquisition unit
6125: Option processing unit
6130: license prediction model value acquisition
6140: licensing prediction information generation unit
6141: Multiple Relationship Processing Module
6142: licensing prediction information value generation module
6143: weight control unit
6150: licensing prediction information provider
6200: Licensing DB Department
6210: Dispute Generation Patent DB
6220: Licensing prediction element value DB
6230: licensing prediction model value DB
6300: Licensing Prediction Information Analysis Engine
6310: Patent Set Division
6400: Licensing Prediction Management
6410: Get Dispute Data
6420: Licensing Prediction System Management
6421: licensing prediction information batch generation unit
6430: Licensing Prediction User Management
6431: Licensing UI
6432: User input patent group management unit
6500: Licensing Prediction Model Generation Engine
6510: licensing prediction element value generator
6520: licensing prediction model generation unit
6530: license prediction model value generation unit by patent
6540: licensing prediction model value provider
8400: e-commerce service provider
8410: DB for e-commerce
8420: Electronic Commerce Processing Department
8430: Open Market Generator
8500: Community Service Provider
8600: Billing processing unit
8700: Comment processing unit
8710: Comment Receiving Unit
8720: Comment storage unit
8730: Identity identification unit
8740: Comment delivery unit
8800: Translation Processing Unit

Claims (22)

(i) (i1) (i11) first obtain a target patent set comprising at least one or more patents, and (i12) obtain at least one complementary patent set having a predetermined association with the target set, or (i2) (i21 A) first obtaining a complementary patent set including at least one patent, and (i22) obtaining at least one target patent set having a predetermined relationship with the supplement set;
(j) obtaining at least one dispute prediction model value for each patent for each patent constituting the complementary patent set; And
(k) generating at least one risk hedging prediction information by using the dispute prediction model value for each patent; patent risk hedging prediction information processing method of a patent information system comprising a. The method of claim 1,
In the step (i12), the complementary patent set is a prior application based on the best day of the target patents constituting the target patent set, or in the step (i22), the target patent set is the supplement that constitutes the target patent set. A patent risk hedging prediction information processing method of a patent information system, characterized in that it is a post-application based on the best date of the patent. The method of claim 1,
The target patent set is
A patent set having a predetermined relationship with a self patent set including at least one patent, or a partial patent set of a patent set having a predetermined relationship with the self patent set,
A patent set consisting of patents having a preset dispute prediction model value or more, or a partial patent set of a patent set consisting of patents having a predetermined dispute prediction model value or more,
A patent set consisting of patents having at least a predetermined dispute prediction information value in relation to the self patent set, or a portion of a patent set consisting of patents having at least a predetermined dispute prediction information value in relationship with the own patent set. Patent set
A patent set having at least a predetermined dispute prediction information value in relation to the own patent set or a partial patent set of a patent set having at least a predetermined dispute prediction information value in relation to the own patent set Patent risk hedging prediction information processing method of information system. The method of claim 1,
The target patent and the complementary patent is a related patent relationship consisting of patents having a predetermined relationship,
The preset relation may be at least one of a citation-cited patent relationship, a similar patent group relationship in a text mining technique, and a similar technology patent relationship in a patent classification.
Complementary patents constituting the complementary patent set are those that can have a predetermined relationship with two or more target patents,
The relevance can be quantified, wherein the quantification is configured to include any one or more of the relevance frequency and relevance strength patent risk hedging prediction information processing method of a patent information system. 5. The method of claim 4,
The relevance frequency is the number of target patents having a predetermined relevance corresponding to the supplementary patent, and the relevance strength is one or more of citation-cited patent relationship information, similar patent group relationship information, and similar technology patent relationship information. Patent risk hedging prediction information processing method of a patent information system characterized in that it is generated using. 6. The method of claim 5,
Generating the relevance strength with the citation-cited patent relationship information includes generating using one or more of citation-cited depth information and citation-cited type information.
Generating the relevance strength with the similar patent group relationship information is generated by using similarity information between the target patent and the supplementary patent,
Generating the relevance strength with the similar technology patent relationship information may correspond to at least one or more types of patent classifications included in the target patent and the supplementary patent, the matching depth, the matching frequency, and the like in the patent classification system to which the patent classification belongs. Is generated using any one or more of the matching sequences,
The similarity information may include at least one type of extracted keyword extracted from the target patent and the supplementary patent, citation-cited information generated from the reference information of the target patent and the supplementary patent, and at least one kind extracted from the target patent and the supplementary patent. It is generated by including at least one of the patent classification and at least one or more of the higher patent classification of the patent classification in the patent classification system to which the patent classification belongs,
Generating the relevance strength with the similar technology patent relationship information may correspond to at least one or more types of patent classifications included in the target patent and the supplementary patent, the matching depth, the matching frequency, and the like in the patent classification system to which the patent classification belongs. A patent risk hedging prediction information processing method of a patent information system, characterized in that it is generated utilizing any one or more of the matching sequence. The method of claim 1,
The generated risk hedging prediction information
Risk hedging prediction information for each target patent corresponding to the complementary set, at least one risk hedging prediction information for each target patent group corresponding to the supplementary set, risk hedging prediction information for the target patent set corresponding to the supplementary set, and at least one or more Patent risk hedging prediction information of the patent information system, characterized in that any one or more of the risk hedging prediction information for each target patent corresponding to the partial complementary set, and at least one or more risk hedging prediction information for each target patent group corresponding to the at least one partial complementary set. Treatment method. 8. The method of claim 7,
The risk hedging prediction information is any one or more of the preset at least one risk hedging prediction information value, at least one licensing prediction analysis information, at least one licensing prediction basis information processing of the patent risk hedging prediction information of the patent information system Way. The method of claim 1,
Acquiring the complementary patent set in the step (i12) is to obtain only a patent that meets a predetermined condition as a supplementary patent,
Acquiring the complementary patent set in the step (i12) is to obtain the divided partial target patent set, or
Acquiring the complementary patent set of the step (i12) is to obtain for each of the two or more target patent set,
The preset condition is set by the user or set by the patent information system, and the preset condition is any one or more of an owner condition, a relevance condition, an owner's sales amount, an owner's recent dispute complaint information, and an individual patent attribute condition. Will be
When the acquisition of the supplementary patent set is to be obtained for each of the divided partial target patent sets, generating the risk hedging prediction information in the step (k) may be a partial supplement patent set corresponding to the partial target patent set. Is generated by
When the acquisition of the supplementary patent set is obtained for each of the two or more target patent sets, the acquisition of the target patent set in the step (i) is that of the at least two or more target patent sets, wherein the (k) step And generating the risk hedging prediction information is generated for each complementary patent set corresponding to two or more target patent sets. The method of claim 9,
The predetermined condition in the acquisition of the supplementary patent set is that the patent information system or at least one right attribute attribute condition that the supplementary patent selected or set by the user must satisfy is satisfied,
The rights holder attribute information is any one or more of owner size information, rights holder type information, and whether the user is included in at least one group of related rights holders selected or designated by the user. Way. The method of claim 1,
Obtaining the target patent set in the step (i22) is to obtain only a patent that meets a predetermined condition as a target patent,
Obtaining the target patent set in the step (i22) is to obtain by the divided partial complementary patent set, or
Obtaining the target patent set in the step (i22) is to obtain for each of two or more complementary patent set,
The preset condition is set by the user or set by the patent information system, and the preset condition is any one or more of an owner condition, a relevance condition, an owner's sales amount, an owner's recent dispute complaint information, and an individual patent attribute condition. Will be
When the acquisition of the target patent set is to be obtained for each of the divided partial complementary patent sets, the generating of the risk hedging prediction information in the step (k) may include partial target patent sets corresponding to the partial supplement patent sets. Is generated by
When the acquisition of the target patent set is obtained for each of two or more complementary patent sets, the acquisition of the supplementary patent set in step (i) is that of at least two or more complementary patent sets, and the (k) step Generating the risk hedging prediction information is generated for each target patent set corresponding to two or more of the complementary patent sets. The method of claim 1,
At least one of a target patent constituting the target patent set and at least one or more partial target patent sets to which at least one or more division criteria or at least one or more selection criteria are applied has a predetermined ranking,
And said ranking is generated according to a ranking generation rule using at least one of said association frequency, said association strength, and said dispute prediction model value. The method of claim 1,
Generating the risk hedging prediction information of step (k) is generated by at least two divided or selected complementary patent sets,
The generation of the risk hedging prediction information in the step (k) is to be generated by reflecting the user weight information received from the user,
The selection or division of the complementary patent set is one or more of a method of selecting or dividing by the user or a method of dividing by a predetermined selection criterion or a division criterion of the system at all times. Or the dividing criterion is at least one of the name of the holder of the supplementary patent, the attribute of the owner, the classification of the patent technology, and at least one predetermined evaluation value constituting the supplementary patent set.
The user weight information includes weight information for each supplementary patent set by the user for each supplementary patent with respect to any one or more of the supplementary patents constituting the supplementary patent set, weight information for each attribute set in an individual attribute of a target patent, and citation- Citation-cited weight information set for the citation relationship, text mining weight information set for the text mining relationship, and similar technology patent relationship weight information set for the similar technology patent relationship in the patent classification. Patent risk hedging prediction information processing method of a patent information system. The method of claim 1,
Obtaining the dispute prediction model value for each patent in the step (j) is performed on the secondary supplementary patents constituting the secondary supplementary patent set generated after performing a predetermined set operation on the obtained primary supplementary patent set. Will be
The secondary supplementary patent set is performed by one or more of deletion of the primary supplementary patent constituting the primary supplementary patent set, limitation on the primary supplementary patent set, and addition of new patents to the secondary supplementary patent set. And a patent risk hedging prediction information processing method of a patent information system, characterized in that it is generated. The method of claim 1,
The dispute prediction model value for each patent is
Any one or more disputes selected from at least one citation perspective dispute prediction element, at least one multi-conflict patent perspective dispute prediction element, at least one multi-dispute cause perspective dispute prediction element and at least one multi-conflict technology family viewpoint dispute prediction element A method for processing patent risk hedging prediction information of a patent information system, characterized in that it is generated using a dispute prediction element value of the prediction element. The method of claim 1,
The dispute prediction model value may be assigned to a licensed prediction element value as an explanatory variable value for a licensed patent and a non-licensed patent, and is different from the licensed patent grant value and the licensed patent grant value granted to the licensed patent. And a non-licensing patent grant value assigned as a response variable value to generate a predetermined statistical processing, wherein the patent risk hedging prediction information processing method of the patent information system is generated. 17. The method of claim 16,
The statistical processing is a patent risk hedging prediction information processing method of a patent information system, characterized in that for using the algorithm of the machine learning series using an ensemble technique using a tree. A target patent set generation unit for obtaining or generating a target patent set including at least one patent;
A supplementary patent set generating unit for obtaining or generating a supplementary patent set including at least one patent;
A dispute prediction model value obtaining unit which obtains at least one dispute prediction model value for each patent for the individual patents constituting the supplementary patent set; And
And a risk prediction information generation unit for generating at least one risk hedging prediction information by using the dispute prediction model value for each patent. 19. The method of claim 18,
And a multiple relationship processing module for calculating a predetermined relationship between a target patent constituting the target patent set and a supplementary patent constituting the supplementary patent set.
The predetermined relationship calculated by the multi-relationship processing module is any one or more of a citation-cited patent relationship, a similar patent group relationship in a text mining technique, and a similar technology patent relationship in a patent classification. Patent information system to process. A patent information system for processing patent risk hedging prediction information, comprising performing the method of any one of claims 1 to 17. 18. A recording medium having recorded thereon a computer readable program for performing the method of any one of claims 1 to 17. delete

Images