JP7303614B2 - making device - Google Patents

Description

The present invention relates to a production device .

Although there are some differences depending on the country, a patent application basically consists of a specification, claims, an abstract, and drawings. Traditionally, the claims (also referred to as claims) of a patent application have been drafted by patent attorneys or by members of a company's intellectual property department.

In recent years, the term "artificial intelligence (AI)" has become common (see, for example, Non-Patent Document 1). In fact, AI has evolved remarkably, and some have a cognitive function that recognizes many images and sounds and finds specific patterns, like the human brain. For example, Google succeeded in getting computers to recognize images of cats. As the performance of AI improves in this way, its application range is also expanding.

http://www.nri.com/jp/journal/2016/10/161031_4/, retrieved on April 10, 2018

In the conventional complaint preparation method, the quality of the prepared complaints often varied because the knowledge and experience required for preparing complaints differed depending on the person in charge. In addition, it took a relatively long time because it was created manually.

SUMMARY OF THE INVENTION The present invention has been made in view of these problems, and its object is to provide a technique capable of suppressing variations in quality when creating claims or reducing the time required to create claims.

One aspect of the present invention relates to a creation device. This drafting device includes receiving means for receiving information, and a plurality of claim drafting models corresponding to different claim structures, each of which is extracted from a past application and machine-learning the scope of claims of that application. applying means for applying the received information to a model selected based on user input from among the plurality of claim creation models , which are models generated by and providing means for providing.

It should be noted that any combination of the above constituent elements, or any mutual replacement of the constituent elements or expressions of the present invention with devices, methods, systems, computer programs, recording media storing computer programs, etc., also constitutes the present invention. It is effective as an aspect of

Advantageous Effects of Invention According to the present invention, it is possible to suppress variations in quality when writing a claim, or to reduce the time required to write a claim.

FIG. 4 is a schematic diagram showing transition of screens realized by the creation server according to the first embodiment; 1 is a schematic diagram showing the configuration of a claim drafting support system according to a first embodiment; FIG. 3 is a hardware configuration diagram of a creation server in FIG. 2; FIG. FIG. 3 is a block diagram showing the functions and configuration of the creation server of FIG. 2; 5 is a data structure diagram showing an example of a patent information holding unit in FIG. 4; FIG. FIG. 3 is a flow chart showing a series of processes in the creation server of FIG. 2; FIG. FIG. 4 is a schematic diagram showing an example of the relationship between the free keywords of the document to be learned and the free keywords of the cited document. FIG. 11 is a flow chart showing the flow of keyword extraction processing according to the first modified example; FIG. It is a representative screen figure of an extraction result display screen. FIG. 4 is a schematic diagram for explaining a claim generation model when designating a model application; FIG. 11 is a schematic diagram showing the flow of complaint generation in the generation server according to the second modification; FIG. 11 is a schematic diagram illustrating learning of an AI model in a creation server according to a second modified example; FIG. 14 is a schematic diagram showing the flow of complaint generation in the generation server according to the third modification; FIG. 14 is a schematic diagram showing the flow of complaint generation in the generation server according to the third modification; FIG. 20 is a schematic diagram showing the flow of complaint generation in the generation server according to the fourth modification; FIG. 21 is a schematic diagram illustrating learning of an AI model in a creation server according to a fourth modified example; FIG. 9 is a block diagram showing functions and configuration of a creation support server according to a second embodiment; 17 is a schematic diagram showing an overview of transition of screens displayed on the display of the user terminal by the creation support server of FIG. 16; FIG. FIG. 11 is a representative screen diagram of a self-project list screen; FIG. 11 is a representative screen diagram of a new item creation screen; FIG. 10 is a representative screen diagram of a configuration requirement model screen; FIG. 4 is a schematic diagram for explaining an input support function; It is a representative screen diagram of a new claim creation screen. FIG. 11 is a representative screen diagram of a screen listing a company's own project search; FIG. 11 is a representative screen diagram of a matter detail screen; FIG. 11 is a representative screen diagram of a matter correction screen; It is a representative screen diagram of a claim correction screen. FIG. 17 is a flow chart showing a series of processes in the creation support server of FIG. 16; FIG. FIG. 20 is a data structure diagram showing an example of a constituent table included in the creation support server according to the fifth modified example; 1 is a schematic diagram showing an exemplary class component model; FIG. FIG. 10 is a schematic diagram showing the main part of the case component model after clicking the plus button; FIG. 10 is a schematic diagram showing the main part of the case component model after clicking the plus button; FIG. 30 is a schematic diagram showing a case component model obtained by correcting the case component model of FIG. 29;

Hereinafter, the same or equivalent constituent elements, members, and processes shown in each drawing are denoted by the same reference numerals, and duplication of description will be omitted as appropriate. Also, in each drawing, some members that are not important for explanation are omitted.

(First embodiment)
FIG. 1 is a schematic diagram showing transition of screens realized by the creation server according to the first embodiment. The creation server first causes the display 100 of the user's terminal to display a creation condition input screen 102 for accepting input of conditions for creating a complaint. The generation condition input screen 102 includes a keyword input area 104, an add button 106, an image upload area 105, a claim number input area 108, a problem input area 110, a reference application input area 107, an autobox 109, and fields. It has an input area 112 and a generate button 114 .

A user of the production server is assumed to be, for example, a person in an intellectual property department of a company. A user examines a technical document such as an invention proposal submitted by a research and development department, and extracts the essence (or key point, feature, or essential part) of an invention or technical idea contained therein as a keyword. In addition, if possible, the user extracts problems related to inventions and technical ideas, and determines the technical field to which the inventions and technical ideas belong. The user inputs the extracted keyword into the keyword input area 104 of the generation condition input screen 102 . The user can increase the number of keywords that can be input in the keyword input area 104 by clicking (pressing) the add button 106 . Each keyword in the keyword input area 104 is displayed in association with a degree of importance 103 . The user selects the importance level 103 of the input keyword from "H" (high), "M" (middle), and "L" (low). In other embodiments, more levels of importance may be used, or sequential values may be used.

Instead of or in addition to keywords, users can upload technical documents such as invention proposals, technical drawings, still images such as photographs, and various files such as moving images (hereinafter referred to as documents). The user inputs the path of the target file in the image upload area 105 by referring to a document or the like in the user's terminal. Alternatively, the user may enter the path of the target file directly into the image upload area 105 . Alternatively, the user may upload a file of an image of a technical document such as an invention proposal instead of a document or the like.

The user determines the number of claims to be created and inputs it in the claim number input area 108 . If the subject of the invention and the technical field are determined, the user inputs them in the subject input area 110 and the field input area 112, respectively. In the challenge input area 110, the challenge may be entered in free text form. If the user has decided on a patent application having claims to be used as a model (hereinafter referred to as a model application), the user enters the patent application number in the reference application input area 107 . Alternatively, check the autobox 109 to automatically select the model application. If the model application is not used, leave the reference application input area 107 blank and uncheck the auto box 109 . In the field input area 112, the technical field may be specified by classification such as IPC (International Patent Classification), F-term, FI, CPC, etc., or may be specified by text.

After the user has entered the necessary information, when the create button 114 is clicked, the creation server acquires the entered information. The production server implements a claim generation AI engine that is generated and updated by having the AI learn information from past patent applications. The creation server automatically generates claims corresponding to the input keywords by processing the input information with this claim generation AI engine. The creation server causes the display 100 to display a creation result display screen 116 displaying the created claims. The generation result display screen 116 displays the generated claims in text, and also displays an OK button 118 and a correction button 120 .

If the user is satisfied with the generated claims, he clicks the OK button 118 to end the process. If the user wishes to modify the generated claims, he clicks on the modify button 120 . When the correction button 120 is clicked, the creation server causes the display 100 to display an edit screen 122 for receiving corrections of the scope of claims by the user.

On the edit screen 122, the claims are displayed in an editable form, and a reflect button 124 is displayed. Modifications made to the scope of claims on the edit screen 122 are displayed in a form that allows the revision history to be understood. When the user clicks the reflect button 124 after making the desired corrections, the creation server causes the display 100 to display a correction result display screen 126 displaying the scope of claims reflecting the corrections. At the same time, the creation server acquires the claims before revision and the revised content, and updates the claim generation AI engine with the acquired information.

In this way, according to the complaint creation support service provided by the creation server, since the complaint creation AI engine is used, it is possible to reduce or eliminate individual factors when creating complaints, and it is possible to achieve uniform quality. can. Also, it is possible to reduce the time required to create a complaint.

FIG. 2 is a schematic diagram showing the configuration of the complaint drafting support system 2 according to the first embodiment. The complaint drafting support system 2 includes a drafting server 4 and a user terminal 8 . The production server 4 and the user terminal 8 are communicably connected via a network 6 such as the Internet. The creation server 4 has a patent information holding unit 42 that holds information on past patent applications, and also has a claim creation AI engine as described above. The user terminal 8 is a terminal used by a user, and may be, for example, a desktop PC, a laptop PC, a mobile terminal, or the like.

In this embodiment, the user inputs information to the user terminal 8, the user terminal 8 transmits the information to the production server 4 via the network 6, the production server 4 processes the information, and the processing result is sent to the network. 6 to the user terminal 8, and the user terminal 8 displays the processing result, a so-called ASP (Application Service Provider) is assumed. However, the technical idea of this embodiment can be applied to systems other than ASP, such as when a complaint creation AI engine is mounted on a stand-alone terminal, and the user inputs information to the terminal and obtains the processing result. is.

FIG. 3 is a hardware configuration diagram of the creation server 4 of FIG. The user terminal 8 may have a hardware configuration similar to that shown in FIG. Production server 4 includes memory 130 , processor 132 , communication interface 134 , display 136 and input interface 138 . These elements are each connected to bus 140 and communicate with each other via bus 140 .

The memory 130 is a storage area for storing data and programs. Data and programs may be stored in memory 130 permanently or temporarily. In particular, memory 130 has an area allocated as patent information holding section 42 . The processor 132 realizes various functions in the creation server 4 by executing programs stored in the memory 130 . The communication interface 134 is an interface for transmitting and receiving data to and from the outside of the creation server 4 . For example, communication interface 134 includes an interface for accessing network 6 . The display 136 is a device for displaying various information, such as a liquid crystal display or an organic EL (Electroluminescence) display. The input interface 138 is a device for receiving input from the user. The input interface 138 includes, for example, a mouse, a keyboard, and a touch panel provided on the display 138 .

FIG. 4 is a block diagram showing the functions and configuration of the creation server 4 of FIG. Each block shown here can be implemented by hardware such as a computer CPU and other elements and mechanical devices, and is implemented by software such as a computer program. It depicts the function blocks to be used. Therefore, those skilled in the art who have read this specification will understand that these functional blocks can be implemented in various ways by combining hardware and software.

The creation server 4 includes a patent information holding unit 42 , a learning unit 402 , a generation condition acquisition unit 404 , a complaint generation AI engine 406 , a result output unit 408 and a correction reception unit 410 .

FIG. 5 is a data structure diagram showing an example of the patent information holding unit 42. As shown in FIG. The patent information holding unit 42 holds information on past patent applications. The patent information holding unit 42 stores a patent ID that specifies a patent application, an application number of the patent application, a leading IPC attached to the patent application, a patent strength value of the patent application, and a patent strength value attached to the patent application. The examiner's free words, the free keywords attached to the patent application, the claims of the patent application, the problems of the patent application, the literature cited during the examination of the patent application, and the The drawing (not shown) of the patent application and the document etc. (not shown) of the patent application are stored in association with each other. Documents such as patent applications may be registered only if they are available, such as documents related to invention proposals filed earlier by the company.

The patent strength value is one of the indicators of the quality of patent applications. The patent strength value may be, for example, a TS (Technology Size) value provided by NRI Cyber Patent Co., Ltd. Alternatively, other indicators may be used.

A free keyword is a keyword set by a person reading the target patent application, and may be a keyword particularly extracted by the person as a point of the target patent application. A free keyword may be, for example, a PATOLIS keyword provided by NRI Cyber Patent Co., Ltd.

Returning to FIG. 4, the learning unit 402 refers to the patent information holding unit 42 and mechanically learns the information extracted from past patent applications and the claims of the patent applications to create a claim creation model. Generate. At this time, the learning unit 402 may not refer to past patent application information that does not meet a predetermined quality standard among the information on past patent applications held in the patent information holding unit 42 .

The learning unit 402 may identify a patent application having a patent strength value equal to or greater than a threshold among the patent applications held in the patent information holding unit 42 . In other words, the learning unit 402 may exclude patent applications whose patent strength value is less than the threshold value from learning targets. In the present embodiment, a magnitude relationship with a threshold value is used as a predetermined criterion for quality, but the present invention is not limited to this. For example, a range of patent strength values may be used instead of the threshold value.

The learning unit 402 acquires the free keywords, claims, problems, leading IPC, documents, etc., and drawings of the specified patent application from the patent information holding unit 42, and mechanically learns the acquired information. , to generate a claim-making model. Generation of the claim creation model in the learning unit 402 may be realized using known machine learning and artificial intelligence technology. Both learning the task and learning the lead IPC are optional and need not be the subject of learning.

The claim creation model is a model that takes keywords or documents as mandatory inputs, issues and technical fields as optional inputs, and outputs claims as a result of processing. Both issues and technical areas are used to improve the accuracy of the output results. When the importance level specified for each keyword is used as an input for the complaint generation model, the learning unit 402 learns free keywords together with their importance levels. The degree of importance of a free keyword may be determined by a person in advance and registered in the patent information holding unit 42, or may be automatically assigned as implemented by a keyword extraction model described later.

FIG. 10 is a schematic diagram for explaining a claim generation model when designating a model application. The learning unit 402 first divides the information of past patent applications held in the patent information holding unit 42 according to the claim structure. Specifically, the learning unit 402 divides the claim structure into several types (for example, claim structures A, B, and C), and determines which type of claim structure of the patent application held in the patent information holding unit 42. determine whether it belongs to The learning unit 402 groups the patent applications held in the patent information holding unit 42 according to this type of structure (for example, claim structure: group belonging to A, claim structure: group belonging to B, claim structure: belonging to C group). The learning unit 402 generates a complaint creation model for each group (that is, for each type of complaint structure). As a result, for example, a claim drafting model A corresponding to the claim structure A, a claim drafting model B corresponding to the claim structure B, and a claim drafting model C corresponding to the claim structure C are generated.

Returning to FIG. 4 , the learning unit 402 sets the complaint generation model generated as described above in the complaint generation AI engine 406 . The type of complaint generation model generated by the learning unit 402 may be determined according to what kind of input is assumed as conditions for complaint generation, or a plurality of types of complaints may be generated according to the assumed input pattern. A creation model may be generated. For example, if a keyword is required as a condition for creating a claim and an issue is optional, the learning unit 402 acquires the free keyword, the scope of claims, and the issue of the specified patent application from the patent information holding unit 42. Then, by mechanically learning the acquired information, a claim generation model is generated. Alternatively, for example, if a keyword is required as a condition for creating a claim, and the degree of importance is optional, the learning unit 402 learns the specified free keyword of the patent application, the scope of claims, and the degree of importance of each free keyword. are acquired from the patent information holding unit 42, and the acquired information is mechanically learned to generate a claim creation model. Alternatively, for example, if documents, etc., are required as conditions for creating a claim, and optional information is not required, the learning unit 402 may convert the claims, documents, etc., and drawings of the identified patent application into patent information. Acquired from the holding unit 42 and mechanically learning the acquired information, a complaint creation model is generated. Alternatively, if it is possible to specify a model application as a condition for drafting claims, a plurality of models for drafting claims are prepared according to the structure of the claims, as shown in FIG.

The generation condition acquisition unit 404 generates screen information for displaying the generation condition input screen 102 and transmits it to the user terminal 8 via the network 6 . The generation condition acquisition unit 404 receives information input by the user to the generation condition input screen 102 displayed on the display 100 of the user terminal 8 from the user terminal 8 via the network 6 . The information entered by the user includes a set of keywords or documents, etc., the number of claims, and optionally the importance of each keyword, the issue, the application number of the model application, and the technical field.

When the autobox 109 is checked, the generation condition acquisition unit 404 searches an external or internal database of patent applications using the received set of keywords as a key, and selects the one with the highest match score among the search results. Set the higher application as the model application.

The complaint generation AI engine 406 applies the set of keywords or documents received from the user terminal 8 by the generation condition acquisition unit 404 and the number of complaints to the complaint generation model generated by the learning unit 402 . If the generation condition acquisition unit 404 receives the degree of importance of the keyword from the user terminal 8, the complaint generation AI engine 406 also applies it to the complaint generation model. If the generation condition acquisition unit 404 receives a problem from the user terminal 8, the complaint generation AI engine 406 also applies it to the complaint generation model. If the generation condition acquisition unit 404 has received a technical field from the user terminal 8, the complaint generation AI engine 406 also applies it to the complaint generation model. If the generation condition acquisition unit 404 has acquired a model application, the claim generation AI engine 406 selects and uses a claim generation model corresponding to the claim structure of the model application. The claim generation AI engine 406 obtains the claims output by the claim generation model as a result of application.

The result output unit 408 provides the user of the user terminal 8 with the claims obtained by the claim generation AI engine 406 . The result output unit 408 generates screen information for displaying the generated result display screen 116 including the obtained claims, and transmits the screen information to the user terminal 8 via the network 6 . The user terminal 8 receives the screen information and causes the display 100 to display the generation result display screen 116 based on the received screen information.

When the correction button 120 on the generation result display screen 116 is clicked, the user terminal 8 generates correction request information and transmits it to the creation server 4 via the network 6 . Upon receiving the correction request information from the user terminal 8 , the correction receiving section 410 generates screen information for displaying the editing screen 122 and transmits the screen information to the user terminal 8 via the network 6 . The user terminal 8 receives the screen information and causes the display 100 to display the edit screen 122 based on the received screen information.

When the reflection button 124 on the edit screen 122 is clicked, the user terminal 8 generates correction completion information including correction contents at that time, and transmits it to the production server 4 via the network 6 . Upon receiving the correction completion information from the user terminal 8 , the correction reception unit 410 generates screen information for displaying the correction result display screen 126 and transmits the screen information to the user terminal 8 via the network 6 . The user terminal 8 receives the screen information and causes the display 100 to display the correction result display screen 126 based on the received screen information.

Upon receiving the correction completion information, the correction receiving unit 410 obtains (1) a set of keywords input by the user, (2) an input problem (if any), ( Learns 3) input technical fields (if any), (4) claims as generated by the claim generation model (i.e., before revision), and (5) claims after revision 402. The learning unit 402 updates the complaint creation model based on the information (1) to (5) above.

The operation of the creation server 4 configured as above will be described.
FIG. 6 is a flow chart showing a series of processes in the creation server 4 of FIG. In the example of FIG. 6, it is assumed that a keyword and a model application are given as conditions for creating a claim. The creation server 4 displays the creation condition input screen 102 on the display 100 of the user terminal 8 (S602). The creation server 4 receives a set of keywords related to the technical idea and the application number of the model application from the user via the creation condition input screen 102 (S604). The preparation server 4 acquires the claims of the model application from an external or internal patent application database (eg, J-Plat, etc.) based on the application number of the model application, and identifies the structure of the acquired claims (S6051). Creation support server 4 selects a complaint creation model corresponding to the specified structure (S6052). Creation server 4 uses the set of keywords received in step S604 as an input, and executes complaint creation by AI using the complaint creation model selected in step S6052 (S606). The creation server 4 displays the generated complaint by displaying the generation result display screen 116 on the display 100 of the user terminal 8 (S608). If the user does not correct the complaint (NO in S610), the process ends. If the complaint is to be corrected by the user (YES in S610), the production server 4 accepts the details of the correction (S612). The creation server 4 displays the corrected complaint by displaying the correction result display screen 126 on the display 100 of the user terminal 8 (S614). The generation server 4 updates the complaint generation model of the complaint generation AI engine 406 based on the complaint before revision and the content of the revision (S616).

In the above-described embodiments, examples of holding units are hard disks and semiconductor memories. In addition, based on the description of this specification, each part is implemented by a CPU (not shown), an installed application program module, a system program module, a semiconductor memory that temporarily stores the contents of data read from the hard disk, etc. Those skilled in the art who have access to this specification will understand what is possible.

According to the creation server 4 according to the present embodiment, a claim creation model for a claim creation AI engine is generated by mechanically learning information on relatively high-quality past patent applications. Claims are automatically generated by applying keywords extracted by the user to this claim creation model. In this way, by transferring the knowledge and experience of claim drafting, which has traditionally belonged to patent attorneys and intellectual property departments of companies, to AI, and by using AI to draft claims, The quality of claims can be uniformed and the time required for claim drafting can be reduced.

The configuration and operation of the complaint drafting support system 2 according to the first embodiment have been described above. Those skilled in the art will understand that this embodiment is merely an example, and that various modifications are possible in combination of each component and each process, and that such modifications are within the scope of the present invention.

In the first embodiment, the free keyword stored in the patent information holding unit 42 is used as the keyword to be learned by the learning unit 402, but the present invention is not limited to this. For example, examiner free words may be used.

(First modification)
Of the patent applications held in the patent information holding unit 42, AI may be used to set free keywords for patent applications that do not have free keywords. In the first modified example, the learning unit 402 identifies patent applications having a patent strength value equal to or greater than a threshold among the patent applications held in the patent information holding unit 42 . The learning unit 402 mechanically learns the free keywords of the documents cited during the examination of the patent application, in addition to the specified free keywords of the patent application, the scope of claims, and the issues to be addressed. Generate an extraction model.

The relationship between the free keywords of patent applications to be learned and the free keywords of cited documents in machine learning will be explained. FIG. 7 is a schematic diagram showing an example of the relationship between the free keywords of the document to be learned and the free keywords of the cited document. In the example of FIG. 7, a first document 752 and a second document 754 were cited during examination of the document 750 to be studied. In order to indicate this citation relationship, in the patent information storage unit 42, the number of the publication of the first document 752 and the number of the publication of the second document 754 are entered in the column of the cited document corresponding to the patent ID of the document 750 to be learned. and are registered. The learning unit 402 identifies the cited document of the document 750 to be learned by referring to the cited document column. In the example of FIG. 7, the first document 752 has two free keywords "AAA" and "BBB", the second document 754 has three free keywords "AAA", "CCC", and "DDD". There are three free keywords in the document 750 of ``AAA'', ``BBB'', and ``EEE''. These free keywords are all held in the patent information holding unit 42 and can be referenced by the learning unit 402 .

When the learning unit 402 mechanically learns the document 750 to be learned, the learning unit 402 refers to the free keywords of the first document 752 and the second document 754 that are the cited documents of the document 750 to be learned. The learning unit 402 sets the importance of the free keyword of the document 750 to be learned based on the free keyword of the cited document. In particular, the learning unit 402 sets a higher degree of importance for a cited document that is less overlapped with the free keyword. For example, the free keyword "AAA" of the document 750 to be learned is also the free keyword of both the first document 752 and the second document 754, so it is assigned the lowest importance. The free keyword "EEE" of the document 750 to be learned is not the free keyword of either the first document 752 or the second document 754, so it is assigned the highest importance. The free keyword "BBB" of the document 750 to be learned is the free keyword of the first document 752 but not the free keyword of the second document 754, so it is assigned an intermediate degree of importance. The learning unit 402 learns the document 750 to be learned, including the degree of importance set in this way, and generates a keyword extraction model.

FIG. 8 is a flowchart showing the flow of keyword extraction processing according to the first modification. The creation server 4 accepts patent applications with no free keywords among the patent applications held in the patent information holding unit 42 as documents targeted for keyword extraction (S702). The creation server 4 uses the information (body text, etc.) of the target document received in step S702 as an input, and executes extraction of constituent elements (keywords) by AI (S704). In particular, the creation server 4 applies the information of the target document received in step S702 to the keyword extraction model generated by the learning unit 402, and acquires the keywords output by the keyword extraction model. The creation server 4 displays the extracted keyword on the display 100 of the user terminal 8 (S706). If the keyword is not corrected by the user (NO in S708), the creation server 4 registers the extracted keyword in the patent information holding unit 42 as a free keyword corresponding to the target document, and terminates the process.

If there is a keyword correction by the user (YES in S708), the creation server 4 accepts the content of the correction (S710). The creation server 4 updates the keyword extraction model based on the keyword before correction and the contents of correction (S712). The production server 4 registers the corrected keyword in the patent information holding unit 42 as a free keyword corresponding to the target document, and terminates the process.

The creation server 4 causes the display 100 to display an extraction result display screen 902 displaying the extracted keywords in step S706. FIG. 9 is a representative screen diagram of the extraction result display screen 902. As shown in FIG. An extracted keyword is displayed on the extraction result display screen 902, and an OK button 118 and a correction button 120 are also displayed. The keywords displayed on the extraction result display screen 902 are ranked according to importance levels of "high", "middle", and "low".

In the above-described first modification, a case has been described in which a keyword extraction model is used to extract and register appropriate keywords for patent applications to which free keywords are not attached, but the present invention is not limited to this. For example, the creation server 4 may accept a document uploaded by a user as a document targeted for keyword extraction. In this case, the user can check the keyword extracted from the document uploaded by the user, together with the rank thereof, by viewing the extraction result display screen 902 shown in FIG.

In the above-described first modification, the case where a keyword extraction model is generated from the information held in the patent information holding unit 42 has been described. may be generated. Similar to patent documents, academic papers are associated with issues, hand-selected keywords, abstracts corresponding to claims, and citation/citation relationships between papers.

(Second modification)
In the first embodiment, the case where the user is requested to input a keyword or an image file has been described, but the present invention is not limited to this. The creation server 4 according to the second modification acquires technical documents themselves such as invention proposals from users, and uses a keyword extraction model, a technical classification model, a similar document extraction model, and a claim creation model. Automatically capture the required information and generate a claim. In addition, in this modified example, the case of using the technical classification assignment model will be explained, but it is not limited to this. information may be used. In this modified example, a case where keywords are extracted and used will be described, but this is not the only option, and instead of keywords, a subject-verb pair or a sentence containing a subject and a verb may be extracted and used.

FIG. 11 is a schematic diagram showing the flow of complaint generation in the creation server 4 according to the second modification. A user 302 uploads an invention proposal 304 from the user terminal 8 to the creation server 4 . The creation server 4 applies an automatic keyword extraction AI model 306 and an automatic technology classification AI model 308 to the received invention proposal 304 . The keyword automatic extraction AI model 306 outputs the keyword 310 necessary for the claim, and the technical classification automatic assignment AI model 308 outputs the technical classification. The user 302 confirms the keyword 310 and technical classification 312 through the display 100 of the user terminal 8, and corrects them through the input means (mouse, keyboard, touch panel, etc.) of the user terminal 8 if necessary. The production server 4 accepts corrections to the keyword 310 and generates a corrected keyword 314 . Note that this correction step may be omitted. At this time, the creation server 4 may allow the user 302 to designate the degree of importance for each keyword, and obtain the designated degree of importance. The creation server 4 passes the set of the uncorrected keyword 310 and the corrected keyword 314 to the later-described learning phase for updating the keyword automatic extraction AI model 306 . Note that the creation server 4 may pass only the uncorrected keyword 310 or only the corrected keyword 314 to the learning phase described below for updating the keyword automatic extraction AI model 306 . The production server 4 accepts modifications to the technical classification 312 and generates a modified technical classification 316 . At this time, the creation server 4 may allow the user 302 to designate the degree of importance for each technical classification, and acquire the designated degree of importance. The creation server 4 passes the set of the uncorrected technology classification 312 and the corrected technology classification 316 to the learning phase described later for updating the technology classification automatic assignment AI model 308 .

The production server 4 applies the similar document automatic extraction AI model 318 to the corrected keyword 314 and the corrected technical classification 316 . The similar document automatic extraction AI model 318 outputs similar document 320 . The user 302 confirms the similar document 320 via the display 100 of the user terminal 8 and corrects it via the input means of the user terminal 8 if necessary. Note that this step of confirming and correcting similar documents may be omitted. The creation server 4 accepts corrections to the similar document 320 and generates a corrected similar document 322 . The creation server 4 passes the set of the uncorrected similar document 320 and the corrected similar document 322 to the learning phase described below for updating the similar document automatic extraction AI model 318 . Note that the creation server 4 may pass only the uncorrected similar document 320 or only the corrected similar document 322 to the learning phase described below for updating the similar document automatic extraction AI model 318 .

The production server 4 applies the claim auto-generation AI model 324 to the corrected keywords 314 , the corrected technology taxonomy 316 and the corrected similar documents 322 . Note that the creation server 4 may apply the claim automatic generation AI model 324 to the keyword before correction, the technical classification before correction, and the similar document before correction. Complaint auto-generation AI model 324 outputs complaint 326 . The user 302 confirms the complaint 326 through the display 100 of the user terminal 8 and corrects it through the input means of the user terminal 8 if necessary. Note that the step of amending this claim may be omitted. Production server 4 accepts corrections to claim 326 and generates corrected claim 328 . The production server 4 passes the set of uncorrected claims 326 and corrected claims 328 to the learning phase described below for updating the claim auto-generation AI model 324 .

FIG. 12 is a schematic diagram for explaining AI model learning in the creation server 4 according to the second modification. The creation server 4 includes an automatic keyword extraction learning AI330, an automatic technology classification assignment learning AI332, a similar document automatic extraction learning AI334, and an automatic complaint generation learning AI336.

The keyword automatic extraction learning AI 330 refers to the patent information holding unit 42 and extracts (1) problems, (2) abstracts, (3) abstracts and keywords of documents cited in the examination process, and (4) claims of past patent applications. (Independent Claim) is obtained as an input, and (5) Keywords and (6) Examiner Keywords of the past patent application are obtained as outputs. In addition, the automatic keyword extraction learning AI 330 can be used for past patent applications (1) problems, (2) abstracts, (3) abstracts and keywords of documents cited in the examination process, and (4) claims (independent claims). Some or all of them may be taken as input. At that time, (2) the abstract and (3) the abstract and keywords of the literature cited in the examination process may be obtained from an external commercial database or the like. Also, the keyword automatic extraction learning AI 330 may acquire either (5) keywords or (6) examiner keywords of past patent applications as an output. At that time, (5) the keyword may be obtained from an external commercial database or the like. The keyword automatic extraction learning AI 330 acquires a set of the uncorrected keyword 310 and the corrected keyword 314 passed from the generation phase of FIG. 11 . The keyword automatic extraction learning AI 330 learns based on the acquired information and generates the keyword automatic extraction AI model 306 .

The technology classification automatic assignment learning AI 332 refers to the patent information holding unit 42, acquires (1) problems, (2) solutions, and (3) abstracts of past patent applications as inputs, and ( 4) Take the assigned technology classification as output. Note that the technical classification automatic assignment learning AI 332 may acquire some or all of (1) problem, (2) solution, and (3) summary of past patent applications as input. The technology classification automatic assignment learning AI 332 acquires a set of the technology classification before modification 312 and the modified technology classification 316 passed from the generation phase of FIG. 11 . The technology classification automatic assignment learning AI 332 performs learning based on the acquired information and generates the technology classification automatic assignment AI model 308 .

The similar document automatic extraction learning AI 334 refers to the patent information holding unit 42 and acquires (1) full text, (2) technical classification, (3) keywords, and (4) examiner keywords of past patent applications as inputs, (5) Get the documents cited by the examiner of the past patent application as an output. The similar document automatic extraction learning AI 334 acquires a set of the uncorrected similar document 320 and the corrected similar document 322 passed from the generation phase of FIG. 11 . The similar document automatic extraction learning AI 334 performs learning based on the acquired information and generates a similar document automatic extraction AI model 318 .
Alternatively, instead of the above structure, similar The similar document automatic extraction learning AI may be configured to output the claims or part of the claims.

The claim automatic generation learning AI 336 refers to the patent information holding unit 42, and (1) keywords of past patent applications, (2) examiner keywords, (3) technical classifications, (4) documents cited by the examiner, (5) problems and (6) solutions are obtained as inputs, and (7) claims (independent claims) of the past patent application are obtained as outputs. In addition, the claim automatic generation learning AI 336 can be used for past patent applications (1) keywords, (2) examiner keywords, (3) technical classifications, (4) documents cited by examiners, (5) problems, (6) ) may take some or all of the solutions as input. In addition, the complaint automatic generation learning AI 336 may (1) acquire subject-verb pairs or sentences containing subjects and verbs instead of (1) keywords. The automatic claim generation learning AI 336 obtains a set of pre-correction claims 326 and corrected claims 328 passed from the generation phase of FIG. Complaint auto-generation learning AI 336 learns based on the acquired information and generates complaint auto-generation AI model 324 .

(Third modification)
13a and 13b are schematic diagrams showing the flow of complaint generation in the creation server 4 according to the third modification. The user 302 may upload the invention proposal 304 and the prior patent publication 338 from the user terminal 8 to the creation server 4 . At the same time, the user 302 inputs claim elements (subjects) via the user terminal 8 . The creation server 4 causes the display 100 of the user terminal 8 to display the creation table 340 . The creation server 4 first sets the input claim element in the creation table 340 as it is. The production server 4 automatically replenishes verb candidates for each set claim element.

The user 302 may check the verb candidates by looking at the creation table 340 displayed on the display 100 of the user terminal 8 . The user 302 may make corrections via the input means of the user terminal 8 if necessary. After the creation server 4 receives corrections to the verbs (corrections are optional), it automatically supplements case component candidates for each of the verbs. The user 302 may confirm the candidates for the class component by looking at the creation table 340 displayed on the display 100 of the user terminal 8 . The user 302 may make corrections via the input means of the user terminal 8 if necessary. User 302 may also enter a segmentation.

The creation server 4 sets the input classification as it is after reflecting the correction to the class component. The user 302 may view the preparation table 340 for final confirmation. When the user 302 issues a complaint generation instruction, the creation server 4 automatically connects the words of the case components to generate a case component table. The creation server 4 generates complaints and class component models (described later in the second embodiment) from the generated class component table. User 302 reviews the generated claim and constituent models and modifies them as necessary.

(Fourth modification)
FIG. 14 is a schematic diagram showing the flow of complaint generation in the creation server 4 according to the fourth modification. A user 302 uploads an invention proposal 304 from the user terminal 8 to the creation server 4 . The invention proposal 304 contains the subject matter and features of the invention. At the same time, the user 302 inputs the number of the similar document 342, which is the basic patent, and the number of the similar document 344, which is another similar publication, via the user terminal 8. FIG.

The production server 4 performs a class component analysis of the similar document 342, which is the basic patent (2-1). The creation server 4 extracts from the description of the features of the invention extracted from the invention proposal 304, the "subject to which the verb is dependent", the "verb", and the "case component" (hereinafter referred to as case information) described in the claim. Edit (2-2). The creation server 4 compares the case information edited in (2-2) with the detailed description of the similar document 344, which is a similar publication, and re-edits the case information (2-3). At this time, the creation server 4 re-edits so that the degree of similarity is lowered.

The creation server 4 generates the case information of the invention related to the invention proposal 304 from the case information obtained in (2-1) and the case information obtained in (2-3) (2-4). The creation server 4 responds to three patterns of (1) external addition, (2) internal addition, and (3) replacement at the time of this generation. The creation server 4 generates a claim and a case component model (summary version, detailed version, described later in the second embodiment) from the case information of the invention generated in (2-4) (3). . The user 302 confirms the generated claim and class component model, and corrects/evaluates them as necessary (4).

FIG. 15 is a schematic diagram illustrating AI model learning in the creation server 4 according to the fourth modification. The claim generation model used in the generation server 4 according to the fourth modification includes the problems and features of the invention extracted from the invention proposal 304 at the time of generating the claims, the problems and claims of the similar document 342 as the basic patent, The problems, claims, and detailed explanations of the similar document 344, which is another similar publication, are input, and the claims are output. In the learning phase, the problems and features of the invention extracted from past patent applications, the problems and claims of the primary references cited in the examination process of the patent applications, the problems, claims and detailed explanations of the secondary references, is input and the claims of the patent application are output, a claim formulation model is generated.

(Second embodiment)
In the first embodiment, once the first information (invention proposal, keywords, etc.) is obtained from the user, a complaint is basically generated automatically without user intervention. In the second embodiment, a case of assisting the user in creating complaints will be described. The creation support server according to the second embodiment finds similar prior applications by automatically generated keywords or manually (in the case of automatically generated keywords, the keyword is used as a key to perform a patent search, and the result with the highest degree of matching is , etc.), the found claim of the prior application is displayed graphically, the user manipulates the graphical representation, and the modified graphical representation is returned to the text, thereby assisting the user in creating a new complaint. It should be noted that in another embodiment, from the invention proposal, the figure may be created from the beginning for claim design by oneself.

The configuration of the complaint drafting support system including the drafting support server 14 according to the second embodiment conforms to the configuration shown in FIG.
FIG. 16 is a block diagram showing the functions and configuration of the creation support server 14 according to the second embodiment. Each block shown here can be implemented by hardware such as a computer CPU and other elements and mechanical devices, and is implemented by software such as a computer program. It depicts the function blocks to be used. Therefore, those skilled in the art who have read this specification will understand that these functional blocks can be implemented in various ways by combining hardware and software.

The creation support server 14 automatically generates keywords from the invention proposal received from the user, searches for patent documents using the automatically generated keywords as keys, and identifies the patent document with the highest match score among the search results as the prior document. Then, a UI for creating a new claim based on the identified prior art claim is displayed on the display 100 of the user terminal 8, changes to the elements of the claim are accepted on the UI, and a new claim reflecting the change is written. generated by The creation support server 14 includes a patent information storage unit 42, a case information storage unit 16, a keyword generation unit 420, a prior art acquisition unit 422, a display control unit 424, an input reception unit 426, and a complaint generation unit 428. , provided. The item information holding unit 16 holds information on items generated using the creation support server 14 . Note that the user may newly input a figure to the display 100 without using the prior art document.

The user uploads the invention proposal for which a claim is to be created from the user terminal 8 to the creation support server 14 . The keyword generation unit 420 acquires the uploaded invention proposal, and applies the problems and features of the invention described in the acquired invention proposal to the keyword extraction models according to the first and second modifications. to generate keywords.

The prior document acquisition unit 422 searches an external or internal patent document database using the keyword generated by the keyword generation unit 420 as a key. The prior document acquisition unit 422 acquires the patent document with the highest match score among the search results as the prior document. Alternatively, the prior art document acquisition unit 422 may identify a plurality of patent documents with high matching scores as prior art documents. Alternatively, the prior document acquisition unit 422 may directly receive designation of prior documents from the user (for example, by application number, publication number, patent number, etc.).

The display control unit 424 causes the display 100 of the user terminal 8 to graphically display the claims included in the claims of the prior art document identified by the prior art acquisition unit 422 via the network 6 .
The input reception unit 426 receives changes to the displayed figure via the network 6 . Note that the input reception unit 426 may receive input of a new figure.
A complaint generating unit 428 generates a new complaint in text format based on the changed graphic.

The functions of the display control unit 424, the input reception unit 426, and the complaint generation unit 428 will be described in more detail below with reference to screen transitions. The display control unit 424 causes each screen described later to be displayed on the display 100 of the user terminal 8 via the network 6 . The input reception unit 426 receives correction contents and instructions from the user, which will be described later, from the user terminal 8 via the network 6 . Upon receiving an instruction to generate a claim, which will be described later, the claim generating unit 428 generates a claim in text format based on the modified figure.

FIG. 17 is a schematic diagram showing an overview of transition of screens displayed on the display 100 of the user terminal 8 by the creation support server 14 of FIG. When the user accesses the creation support server 14, a login screen 502 is displayed first. When the user enters authentication information such as an ID and password on the login screen 502, the creation support server 14 performs user authentication. When the user authentication succeeds, the screen displayed on display 100 changes from login screen 502 to own project list screen 504 . The login screen 502 may be implemented using known user authentication techniques.

When the user instructs to create a new item on the own item list screen 504, the screen transitions to the new item creation screen 506. When the user instructs to search for an item, the screen transitions to the own item search list screen 508. When the item number link is specified, the screen responds. The screen transitions to the item detail screen 512 of the item to be processed. When the user inputs the necessary information on the item new creation screen 506 and instructs the creation of claims, the screen changes to a new claims creation screen 510 . When registration is instructed on the new claim creation screen 510 , the screen transitions to a matter details screen 512 .

When the user designates a project number link on the own project search list screen 508 , the screen transitions to the project detail screen 512 of the corresponding project, and when the user designates the IPC field list link, the screen transitions to the IPC field list screen 514 . As the IPC field list screen 514, for example, an IPC field list screen of NRI Cyber Patent Desk 2 provided by NRI Cyber Patent Co., Ltd. may be adopted.

When the user corrects the matter on the matter details screen 512 , the screen transitions to the matter correction screen 516 , and when the matter copy is specified, the screen transitions to the matter new creation screen 506 . When the user modifies the scope of claims on the item modification screen 516 , the screen transitions to the scope of claims modification screen 518 . When an instruction to reflect the correction is made on the claims correction screen 518 , the screen transitions to the matter details screen 512 .

FIG. 18 is a representative screen diagram of the self-issue list screen 504. As shown in FIG. The user inputs information in the search key input area 520 of the self-issue list screen 504 and clicks the search button 522 . The input reception unit 426 acquires the information entered in the search key input area 520, searches the case information holding unit 16 using the acquired information as a key, and searches for a case matching the key among the cases of the company to which the logged-in user belongs. as a search result. The display control unit 424 displays the search results acquired by the input reception unit 426 in the own item list display area 524 of the own item list screen 504 .

When the user clicks the case number link 526 of the case displayed in the own case list display area 524, the screen changes to the case details screen 512 of the case corresponding to the clicked case number link 526. FIG. When the user clicks the matter search button 528 , the screen transitions to the own matter search list screen 508 . When the user clicks the new creation button 530 , the screen transitions to the new item creation screen 506 .

FIG. 19 is a representative screen diagram of the new matter creation screen 506. As shown in FIG. The project new creation screen 506 includes a project number display area 532 for displaying an automatically assigned project number, an invention name input area 534, a prior document input area 536, a proposal upload button 538, and an element model edit screen. It has an area 540 , a model button 542 , a relationship expression generation button 544 , a claim range editing area 546 , and a claim range generation button 548 .

The user enters the title of the invention in the invention title input area 534 . The user enters the prior document number (application number, publication number, patent number, etc.) on which the claim is to be generated in the prior document input area 536 . Alternatively, if the user desires automatic selection of prior art documents, the user can click the Upload Proposal button 538 to display a dialog (not shown) for uploading a file, in which the Invention Proposal file can be uploaded. specify. The user terminal 8 transmits the specified invention proposal file to the creation support server 14 . The creation support server 14 receives the file of the invention proposal, and performs keyword creation, patent document search, and prior document identification as described above. The display control unit 424 causes the identified prior document number to be displayed in the prior document input area 536 .

When the user clicks model button 542, the model diagram is output. When the user clicks the element model automatic generation button 541, the input reception unit 426 acquires the prior art document from the external or internal patent document database using the prior art document number entered in the prior document input area 536 as a key. The display control unit 424 graphically displays the claims of the prior art in the element model editing area 540 based on the flow chart model.

The flow chart model displayed in the element model editing area 540 expresses the overall picture of the claim in the order of input→processing (constituent feature)→output, and includes input/output element 550, constituent feature 552, and others. Requirement 554; The user selects, moves, deletes, adds, etc. the elements of the flowchart model displayed in the element model editing area 540 to generate a new desired model.

FIG. 20 is a representative screen diagram of the configuration requirement model screen 556. As shown in FIG. When the user clicks a configuration requirement 552 displayed in the element model editing area 540, the display control unit 424 causes the display 100 to display a configuration requirement model screen 556 corresponding to the clicked configuration requirement 552 as a separate screen pop-up. The user can define each configuration requirement on the configuration requirement model screen 556 . The configuration requirement model screen 556 has an illustration editing area 558 , a configuration requirement details display area 560 , a generated text display area 562 , and an OK button 564 .

The graphical editing area 558 displays the configuration requirements graphically by dividing them into processing and data. The process flow is indicated by a solid line, and the relationship between data is indicated by a dashed line. The configuration requirement detail display area 560 displays data corresponding to the diagram displayed in the schematic editing area 558 . The configuration requirement detail display area 560 displays the result of breaking down the configuration requirement into "subject", "data", "processing", and "component". When the user performs desired editing on the elements in the graphical editing area 558 , the result of editing is reflected in the configuration requirement detail display area 560 . At this time, "relevance" and "processing content" are automatically estimated and automatically supplemented using a probabilistic language model (Long Short-Term Memory, LSTM). After that, it is also possible for the user to manually correct it. At this time, candidates for other items are also presented using the probabilistic language model. For parts, claim-specific words may be prepared as a dictionary. The generated text display area 562 displays the claim wording (text corresponding to the constituent feature) automatically generated using the data displayed in the constituent feature detail display area 560 . When the OK button 564 is clicked, the screen returns to the item new creation screen 506 .

Returning to FIG. 19 , when the user performs desired editing in the element model editing area 540 and clicks the relationship expression generation button 544 , the edited editing is reflected in the claims editing area 546 . Specifically, the input receiving unit 426 receives changes to the elements of the flowchart model displayed in the element model editing area 540 . The claim generator 428 generates a new skeleton (fixed part) of a claim based on the modified flowchart model. The display control unit 424 displays the generated fixed portion in the claim editing area 546 of the new item creation screen 506 . The complaint generation unit 428 supplements assumed candidates for the input part corresponding to each fixed part from the past appearance frequency information. Here, a probabilistic language model may be used. The user confirms the fixed part and the input part displayed in the claim editing area 546, and if the input part needs to be corrected, it is corrected using the input support function.

FIG. 21 is a schematic diagram for explaining the input support function. When the user clicks one cell 566 in the input section displayed in the claim editing area 546, a plurality of autocomplete candidates 568 selected from past learning data are displayed. When the user selects one of the plurality of autocomplete candidates 568, a further plurality of autocomplete candidates 570 corresponding to the selected autocomplete candidate 568 are displayed. In this way, by continuously suggesting autocomplete candidates, the input can be simplified and the generated claims can be standardized.

Returning to FIG. 19 , when the user clicks the Claims Generation button 548 , the Claims Generation Unit 428 generates a new claim based on the information edited in the Claims Editing Area 546 . This claim is the same claim that would be generated from the flow chart edited in the element model edit area 540 if no edits were made in the claims edit area 546 . The display control unit 424 causes the display 100 to display a new claim creation screen 510 including the generated new claim.

FIG. 22 is a representative screen diagram of the new claim creation screen 510 . The new claim creation screen 510 has a comparison area 572 , a possibility display area 574 , a return button 576 and a register button 578 . A comparison area 572 displays the results of comparing the generated claims and the prior art on a constituent element basis. A comparison area 572 displays the degree of similarity and similarity between the claim and each prior art document for each component. Possibility display area 574 displays the possibility of obtaining a patent numerically. When the user clicks the return button 576 , the screen returns to the new item creation screen 506 . When the user clicks the registration button 578 , the input reception unit 426 associates the case number, the name of the invention, the prior art document, and the generated complaint and registers them in the case information holding unit 16 . At the same time, the display control unit 424 causes the display 100 to display the item details screen 512 showing the registered contents.

FIG. 23 is a diagram of a representative screen of the company case search list screen 508. As shown in FIG. The user enters information into the search key input area 580 on the own project search list screen 508 and clicks the search button 582 . The input reception unit 426 acquires the information input in the search key input area 580, searches the item information storage unit 16 using the acquired information as a key, and generates search results. The display control unit 424 displays the search results generated by the input reception unit 426 in the search result list display area 586 of the company case search list screen 508 . When the user clicks the IPC list button 584 , the screen transitions to the IPC field list screen 514 .

When the user clicks the case number link 588 of the case displayed in the search result list display area 586, the screen changes to the case details screen 512 of the case corresponding to the clicked case number link 588. FIG.

FIG. 24 is a representative screen diagram of the item details screen 512. As shown in FIG. The display control unit 424 refers to the item information holding unit 16 to generate the item details screen 512 and causes the display 100 to display it. The case detail screen 512 includes a case number display area 590 for displaying the case number, an invention name display area 592 for displaying the name of the invention, a prior document display area 594 for displaying the number of the prior document, and a claim flow chart model. , a claims display area 598 for displaying the fixed part and the input part of the claim, a matter copy button 802, and a matter correction button 804. When the user clicks the matter copy button 802 , the screen transitions to a new matter creation screen 506 in which the matter displayed on the matter details screen 512 is copied. When the user clicks the matter correction button 804 , the screen transitions to the matter correction screen 516 for accepting correction of the matter displayed in the matter detail screen 512 .

FIG. 25 is a representative screen diagram of the matter correction screen 516. As shown in FIG. The configuration of the matter correction screen 516 and the flow of the modification processing are the same as the configuration of the new matter creation screen 506 and the flow of the generation processing described with reference to FIG. In the example of FIG. 25 , the user modifies the portion surrounded by the dashed circle in the element model editing area 806 and clicks the relationship expression generation button 808 . Then, the information in the claims editing area 810 is updated, and the user adjusts the input part of the claims editing area 810 as necessary while using the input support function. After adjustment, when the user clicks the Claims Generation button 812 , the Claims Generator 428 modifies the claims based on the information edited in the Claims Editing Area 810 . The display control unit 424 causes the display 100 to display a claim correction screen 518 including the corrected claims.

FIG. 26 is a representative screen diagram of the claim correction screen 518. As shown in FIG. The configuration of the claims correction screen 518 is the same as the configuration of the new claims creation screen 510 described with reference to FIG. When the user clicks back button 814 , the screen returns to item modification screen 516 . When the user clicks the registration button 816, the input reception unit 426 associates the case number, the title of the invention, the prior art document, and the corrected claim and registers them in the case information holding unit 16. FIG. At the same time, the display control unit 424 causes the display 100 to display the item details screen 512 reflecting the correction content.

The operation of the creation support server 14 configured as above will be described.
FIG. 27 is a flow chart showing a series of processes in the creation support server 14 of FIG. The creation support server 14 receives the invention proposal from the user's user terminal 8 via the network 6 (S650). The creation support server 14 generates keywords by applying the keyword extraction model to the received invention proposal (S652). The creation support server 14 searches for patent documents based on the generated keyword (S654). The creation support server 14 identifies prior documents from the search results (S656). In addition to or instead of steps S650, S652, and S654, the creation support server 14 may receive the prior document number from the user terminal 8 in step S656. The preparation support server 14 graphically displays the identified claim of the prior art document on the display 100 of the user terminal 8 (S658). Alternatively, the user may be allowed to draw a new figure from scratch. The creation support server 14 accepts changes made by the user to the displayed graphic (S660). The creation support server 14 generates a new complaint based on the changed graphic (S662).

In the above-described embodiments, examples of holding units are hard disks and semiconductor memories. In addition, based on the description of this specification, each part is implemented by a CPU (not shown), an installed application program module, a system program module, a semiconductor memory that temporarily stores the contents of data read from the hard disk, etc. Those skilled in the art who have access to this specification will understand what is possible.

According to the creation support server 14 according to the present embodiment, the user can create a desired complaint by manipulating a graphic such as a flowchart model. Therefore, it becomes possible to operate more intuitively than directly operating sentences, and user convenience is improved.

Moreover, the preparation support server 14 according to the present embodiment generates a new claim by modifying or changing the identified claim of the prior art document. Therefore, it is possible to shorten the time required to generate a claim compared to generating a claim from scratch. In addition, the quality of complaints can be made uniform.

The configuration and operation of the claim drafting support system according to the second embodiment have been described above. Those skilled in the art will understand that this embodiment is merely an example, and that various modifications are possible in combination of each component and each process, and that such modifications are within the scope of the present invention.

(Fifth modification)
In the second embodiment, a case where a flowchart model is adopted as graphic display has been described, but the present invention is not limited to this. For example, a class component model may be employed as a graphic representation.
FIG. 28 is a data structure diagram showing an example of the constituent table 818 provided in the creation support server 14 according to the fifth modification. The constituent table 818 is generated by constituent analysis of prior art claims or by constituent analysis of new claims. The class component analysis may be implemented using the technique described in Japanese Patent Application Laid-Open No. 2015-072573, for example.

The creation support server 14 classifies the words registered in the case component table 818 into the following three types.
(1) Words necessary for composing a claim Appearing in a claim of a prior art document and being or included in a keyword of the prior art document and appearing in a keyword of a document cited in the examination process of the prior art document .
(2) Words that characterize the claims appear in the claims of the prior art documents and are or are included in the keywords of the prior art documents and do not appear in the keywords of the documents cited during the examination of the prior art documents.
(3) Others

FIG. 29 is a schematic diagram illustrating an exemplary class component model 820. As shown in FIG. The display control unit 424 refers to the class component table 818 to generate the class component model 820 and displays it on the display 100 of the user terminal 8 . A class component model 820 may be displayed in place of the element model editing area 540 of the case new creation screen 506 of FIG. In the class component model 820, subjects are represented by rectangles with downward arrows. For example, the subject “moving body” is displayed as a rectangle 822 . Verbs are displayed as ellipses with downward arrows. For example, the verb “have” is displayed with an oval 824 . Case components are displayed as rectangles with left-pointing arrows. For example, the case component “image processing device” is displayed as a rectangle 826 .

In the case component model 820, words are displayed in different colors depending on the types to which the words belong. For example, the word "moving object" belongs to "(1) words necessary for constructing a claim" and is displayed in blue, and the word "boarding seat" belongs to "(2) words characterizing a claim" and is displayed in red. and the word "has" is displayed in black because it belongs to "(3) other".

The state of the class component model 820 shown in FIG. 29 is the most compact or most folded state. By clicking the plus button 828, the user can display the collapsed and hidden portion.

FIG. 30 is a schematic diagram showing the main part of the class component model 820 after the plus button 828 is clicked. When the plus button is clicked, hidden lower layers and connections of the corresponding element are displayed. A "connection" is a dashed line representing the connection of the parts expressed as "said" in the claims. In the example of FIG. 30, the lower hierarchy hidden in the case component model 820 of FIG. 29 is displayed in a rectangle 832 corresponding to the case component "image processing device". In particular, within rectangle 832, the subject "image capture device" is represented by rectangle 834, the verb "to acquire" is represented by oval 836, and the case component "experience image according to passenger experience" is represented by rectangle 838. is displayed.

The plus button 828 in FIG. 29 is replaced by a minus button 830 in FIG. 30, and when the user clicks the minus button 830 the display returns to the class component model 820 of FIG. When the user clicks the plus button 840 displayed in association with the word "passenger", the lower hierarchy and connections related to the word "passenger" are displayed.

FIG. 31 is a schematic diagram showing the main part of the class component model 820 after the plus button 840 is clicked. Clicking on the plus button 840 associated with the word "passenger" causes a dashed line 842 to connect between the multiple words "passenger" that are "said" in the claims. The plus button 840 in FIG. 30 is replaced by a minus button 844 in FIG. 30 and when the user clicks the minus button 844 the display returns to the class component model 820 of FIG.

Similar to the element model editing area 540 of the item new creation screen 506 of FIG. 19, the case element model 820 is configured to be able to accept editing of the displayed elements. Specifically, the user can make the following three edits to the class component model 820 .
(1) Modification of element wording (2) Deletion of case components (3) Addition of case components

FIG. 32 is a schematic diagram showing a class component model 850 obtained by modifying the class component model 820 of FIG. The user corrects the word "stereo image" to the word "handle part" for the rectangle 846 of the case component "to take a stereo image of the object to be grasped" displayed in the case component model 820 of FIG. The user also deletes the rectangle 848 of the case component “so that the imaging ranges overlap” displayed in the case component model 820 of FIG. As a result, in the case component model 850 shown in FIG. 30 , the rectangle 846 of the case component “to capture a stereo image of the object to be grasped” is replaced with a rectangle 852 of the case component “to image the handle of the object to be grasped”. and the rectangle 848 of the case component “so that the imaging ranges overlap” is no longer displayed.

The creation support server 14 again updates the case component table 818 according to the corrected case component model 850 and generates a claim (sentence). It should be noted that it is possible to correct the wording from either the figure or the generated claims, and always keep them in sync.

Technical ideas according to the first embodiment and related modifications may be expressed by the following items.
(Item 1)
a receiving means for receiving information;
applying means for applying received information to a model generated by mechanically learning information extracted from a past application and the claims of the application;
and a providing means for providing claims resulting from the application.
(Item 2)
further comprising an amendment means for accepting amendments to the provided claims;
2. The creating apparatus according to item 1, wherein the providing means provides the scope of claims in which the received corrections are reflected.
(Item 3)
3. The production device according to item 2, further comprising updating means for updating the model based on the received modifications.
(Item 4)
The reception means receives an assignment,
The model is generated by mechanically learning information extracted from past applications, issues of the applications, and claims of the applications,
4. The creating apparatus according to any one of items 1 to 3, wherein the applying means applies the received information and the task to the model.
(Item 5)
a holding means for holding information of past applications;
5. The creation device according to any one of items 1 to 4, further comprising learning means for generating the model by referring to the holding means.
(Item 6)
6. The creation apparatus according to item 5, wherein the learning means does not refer to information on past applications that do not meet a predetermined quality standard among the information on past applications held in the holding means.
(Item 7)
7. A production device according to any one of items 1 to 6, wherein the information is a keyword or an image.
(Item 8)
receiving information; and
Applying the received information to a model generated by machine-learning information extracted from a previous application and the claims of that application;
providing a claim resulting from the application.
(Item 9)
the ability to receive information;
A function that provides claims obtained by applying received information to a model generated by mechanical learning of information extracted from past applications and the claims of the said applications. A computer program for realizing , on a computer.

Technical ideas according to the second embodiment and related modifications may be expressed by the following items.
(Item 10)
Display control means for graphically displaying the claims included in the claims of the past application on a display;
receiving means for receiving changes to the displayed figure;
A creation support device, comprising a creation means for creating a new claim in text format based on the changed figure.
(Item 11)
The display control means graphically displays the claims based on the flowchart model,
11. The creation support device according to item 10, wherein the receiving means receives changes to elements of a flowchart.
(Item 12)
The display control means graphically displays the claims based on the case component model,
11. The creation support device according to item 10, wherein the reception means receives a change to a class component.
(Item 13)
means for receiving information in technical literature;
Application means for applying information of the received technical document to a model generated by mechanically learning keywords extracted from past applications, problems of the applications, and claims of the applications;
13. The creation support apparatus according to any one of items 10 to 12, further comprising specifying means for specifying the past application using the keyword obtained as a result of application.
(Item 14)
Graphically presenting on a display claims included in the claims of the prior application;
accepting changes to the displayed geometry;
generating a new claim in text form based on the modified figure.
(Item 15)
A function to graphically display the claims included in the claims of the past application on the display;
the ability to accept changes to the displayed figure;
A computer program for causing a computer to implement a function of generating a new claim in a text format based on a modified figure.

2 Complaint Creation Support System 4 Creation Server 6 Network 8 User Terminal.

Claims (3)

a receiving means for receiving information;
A plurality of claim drafting models corresponding to different claim structures, each of which is a model generated by machine learning of information extracted from a previous application and the claims of the previous application. means for applying the received information to one of the models selected based on user input;
and a providing means for providing claims resulting from the application. The receiving means selects a model application based on the user input,
The applying means selects a claim drafting model corresponding to the structure of the claims of the selected model application as a model to which the received information should be applied.
The production device according to claim 1. Claim 1 or 2, wherein each of the plurality of claim creation models is generated using information extracted from patent applications that meet a predetermined quality standard among past patent applications as a target of the machine learning. creation equipment.

Images