JP6735247B2 - Document classification device, document classification method, and document classification program - Google Patents

Description

The present invention relates to a document classification device, a document classification method, and a document classification program for classifying documents.

In recent years, engineers often read various patent documents for their own purpose each year in order to keep up with the latest technological trends. Although an engineer narrows down the patent documents by using a search formula, the number of documents obtained as a result is sometimes enormous, and it is not realistic to read all the documents. Therefore, the user may perform screening of documents that should be read and documents that should not be read from a huge amount of documents, that is, screening.

Therefore, there are various techniques as a method for assisting such screening. For example, a seed document that represents the content to be searched is added to the document set obtained as a search result by patent classification, clustering is performed based on the similarity of document contents, and patent documents displayed in clusters are sequentially screened. A technique for doing so is disclosed (for example, see Patent Document 1). To calculate the degree of similarity, morphological analysis is performed on the document to decompose the terms into terms, and the degree of similarity between the terms is calculated to perform clustering.

Further, in order to efficiently perform the screening after the document search, a method of automatically turning over the contents showing each document in a document unit is disclosed (for example, see Patent Document 2).

Further, the similarity between the document to be classified and the document set to which the classification is added in advance is calculated based on the keyword in the document, and the specified number of documents most similar to the input document are extracted, and the similarity is calculated. Calculate the classification score of the specified number of documents extracted based on the number of added classifications, extract the classifications for which the calculated score is greater than the specified value, and assign it to the documents to be classified There is also a technique to do so (for example, see Patent Document 3).

JP, 2003-157270, A JP, 2004-234430, A JP, 2007-323454, A

By the way, in the case of the above-mentioned Patent Document 2, although automatic turning is performed for each document, the number of documents to be read by the user does not decrease, and the fact that a person performs screening does not change, so the processing load of that person is not changed. Is a big problem.

Further, in the case of Patent Document 1 and Patent Document 3 described above, since a method is used in which morphological analysis is performed on each document and then the degree of similarity is calculated for each of the extracted terms, morphological analysis and enormous numbers are performed. There is a problem that a huge processing load such as calculation of the similarity of terms is applied to the processor.

Therefore, the present invention has been made in view of the above problems, and an object of the present invention is to provide a document classification device, a document classification method, and a document classification program that have a smaller processing load on a person or a processor than those of Patent Documents 1 to 3. And

In order to solve the above-mentioned problems, a document classification device according to an aspect of the present invention includes first document information indicating a plurality of document data searched according to a search formula, and the document data assigned to the document data. A storage unit that stores the first feature information indicating one or more technical features and the corresponding information indicating whether the document data is a desired document for the user in association with each other, Of the second feature information and the first feature information, which is searched for in the second document information indicating the other document data to which the second feature information indicating one or more technical features is added. Based on the degree, the extraction unit for extracting a predetermined number of document data from the plurality of document data, and the corresponding information associated with each of the predetermined number of document data extracted by the extraction unit, other document data, A determination unit that determines whether the document is a desired document for the user and an output unit that outputs the determination result of the determination unit are provided.

Further, a document classification method according to an aspect of the present invention includes first document information indicating a plurality of document data searched according to a search formula, and one or more technical data of the document data added to the document data. A storage step of storing first characteristic information indicating a characteristic and corresponding information indicating whether or not the document data is a document desired by the user in association with each other, and a new search is performed according to the search formula, Based on the acquisition step of acquiring the second document information indicating the other document data to which the second characteristic information indicating the above technical characteristics is added, and the degree of coincidence between the second characteristic information and the first characteristic information, a plurality of Other document data based on the extraction step of extracting a predetermined number of document data from the document data and the corresponding information associated with each of the predetermined number of document data extracted in the extraction step. It includes a judgment step of judging whether or not there is, and an output step of outputting a judgment result in the judgment step.

Further, the document classification program according to one aspect of the present invention is configured such that, in a computer, one or more of first document information indicating a plurality of document data searched according to a search formula and the document data attached to the document data. The first feature information indicating the technical features of the document and the corresponding information indicating whether or not the document data is a desired document for the user are stored in association with each other, and a new search is performed according to the search formula. Based on an acquisition function of acquiring second document information indicating other document data to which second characteristic information indicating one or more technical characteristics is added, and a degree of coincidence between the second characteristic information and the first characteristic information Then, based on the extraction function of extracting a predetermined number of document data from a plurality of document data and the corresponding information associated with each of the predetermined number of document data extracted by the extraction function, other document data is desired by the user. The determination function for determining whether or not the document is a document and the output function for outputting the determination result of the determination function are realized.

Further, in the above document classification device, the extraction unit extracts a predetermined number from a plurality of pieces of document data in which the degree of matching between the second feature information and the first feature information is high, and the determination unit extracts the extraction unit. When the relevant information associated with the document has more than the threshold value indicating that the document is the document desired by the user, the other document data is determined to be the document desired by the user, and the document desired by the user is determined. If the number of documents indicating that the document is not a document is larger than the threshold value, it may be determined that the other document data is not a document desired by the user.

In the document classification device, the determination unit weights the relevant information associated with the document extracted by the extraction unit according to the degree of coincidence, and based on the relevant information after weighting. Alternatively, it may be determined whether other document data is a document desired by the user.

Further, in the above document classification device, the determination unit may determine the first feature of the document data that is associated with the first feature information and that indicates that the relevant information is a document that the user does not desire. In documents in which the user does not desire other document data having second feature information in which the non-correspondence rate indicating the ratio to the entire document data with which the information is associated matches the first feature information that exceeds the first threshold value. It may be determined that there is.

Further, in the above document classification device, the determination unit determines the first feature of the document data that is associated with the first feature information and that indicates that the relevant information is a document desired by the user. It is a document in which the user desires other document data having second feature information in which the corresponding rate indicating the ratio to the entire document data with which the information is associated exceeds the second threshold value. It may be determined that.

Further, in the document classification device, the first threshold may be larger than the second threshold.

Further, in the above document classification apparatus, when the feature information is used in the search formula, the determination unit makes the determination based on the first feature information excluding the feature information and the second feature information. It may be that.

In the document classification device according to an aspect of the present invention, whether or not other newly searched document data is a desired document for the user is given to the documents searched in advance by the same search formula. The determination can be made based on the degree of coincidence of the characteristic information indicating the technical information and the corresponding information indicating whether or not the document is a document desired by the user. Therefore, the document classification device can screen documents that the user does not have to read, and thus can reduce the number of documents that the user should read. Therefore, the processing load on the user can be reduced. Further, since the document classification device can screen documents without scrutinizing the inside of the documents, the processing load on the document classification device can be reduced.

It is a block diagram which shows the structural example of a document classification device. It is a block diagram which shows the detailed structural example of a document classification device. (A) is a data conceptual diagram which shows an example of the past list which shows the classified list. (B) is a data conceptual diagram showing an example of a new document list showing a list of new documents. It is a data conceptual diagram which shows an example of the matching degree table which sorted the past document by the matching degree with a new document. It is a data conceptual diagram which shows an example of IPC applicable information which shows whether it is noise or it corresponds for every classification. 7 is a flowchart showing an operation of the document classification device and a process relating to advance preparation. 7 is a flowchart showing a process following FIG. 6. 7 is a flowchart showing an operation of the document classification device and showing a process for classifying a new document. It is a flowchart which shows the process following FIG. 10 is a flowchart showing a process following FIG. 9. It is a block diagram which shows the structural example of a document classification device.

Hereinafter, a document classification device according to one embodiment of the present invention will be described in detail with reference to the drawings.

<Embodiment>
<Structure>
FIG. 1 is a block diagram showing a configuration example of a document classification device. As shown in FIG. 1, the storage unit 130, the acquisition unit 110, the extraction unit 121, the determination unit 122, and the output unit 140 are provided.

The storage unit 130 includes first document information indicating a plurality of document data searched according to the search formula, and first characteristic information indicating one or more technical characteristics of the document data added to the document data. The document data is stored in association with corresponding information indicating whether or not the document data is a document desired by the user. The storage unit 130 can be realized by, for example, an HDD (Hard Disc Drive), an SSD (Solid State Drive), a flash memory, or the like, but is not limited to this. Here, the first document information may be data indicating the document itself of the first document information or identification information indicating the first document information. Here, the first characteristic information indicates a technical characteristic of each document data, and examples thereof include IPC, CPC, ECLA, ICO, USC, FI, and F term, but are not limited thereto. Not a thing. Further, the relevant information may be information indicating whether or not each of the documents obtained as a result of the search by visually observing the documents obtained by the search by the user who has performed the search is a desired document for the user. For example, the information may be information such as “corresponding” or “noise”, or information indicating the content such as “neither” or “unknown”. The relevant information may be given by the document classification device. As described above, at least one technical feature is associated with each first document information as the first feature information indicating the technical feature.

The acquisition unit 110 is newly searched according to the search formula and acquires the second document information indicating other document data to which the second feature information indicating one or more technical features is added. The acquisition unit 121 may acquire the second document information by wired or wireless communication, or may acquire other document data stored in advance in the storage unit 130, for example. Alternatively, another storage medium that stores other document data may be connected to the document classification device to obtain the other document data from the other storage medium. The other storage medium is, for example, a portable storage medium such as a flash memory. The second document information may be data indicating the document itself of the second document information or identification information indicating the second document information. The second characteristic information indicates technical characteristics of other document data, and examples thereof include IPC, CPC, ECLA, ICO, USC, FI, F term, but are not limited thereto. Not a thing. Here, the other document data acquired by the acquisition unit 110 may be one, a plurality, or either.

The extraction unit 121 extracts a predetermined number of document data from the plurality of document data based on the degree of coincidence between the second characteristic information and the first characteristic information. The extraction unit 121 can be realized by, for example, a processor that executes the extraction program stored in the storage unit 130. For example, the extraction unit 121 may extract the document data in which the degree of coincidence between the second characteristic information and the first characteristic information is high, from the plurality of document data, or the document data in which the degree of coincidence is equal to or higher than a certain level. It may be done.

The determination unit 122 determines whether other document data is a desired document for the user, based on the corresponding information associated with each of the predetermined number of document data extracted by the extraction unit 121. The determination unit 122 can be realized by, for example, a processor that executes the determination program stored in the storage unit 130. For example, when the relevant information associated with the document data extracted by the extracting unit 121 indicates a lot of relevant information, the determining unit 122 may determine that the other document data also corresponds to the document desired by the user. it can. It should be noted that the condition "is the document desired by the user?" is a condition "is the document satisfying a predetermined viewpoint?" or the condition "is the document satisfying the predetermined condition?" Good.

The output unit 140 outputs the determination result of the determination unit 122. The output unit 140 only needs to be able to output the determination result of the determination unit 122 to the outside. For example, the document classification device 100 includes a monitor or a speaker as an output device and outputs the determination result as image information to these monitors, or The determination result may be output as voice information. Further, the output unit 140 may be configured to output by transmitting an information indicating the determination result to the external device wirelessly or by wire, for example, when an external device is connected to the document classification device 100.

Hereinafter, the document classification device 100 will be described in more detail.

FIG. 2 is a block diagram showing a detailed configuration example of the document classification device 100. As shown in FIG. 2, the document classification device 100 includes an acquisition unit 110, a control unit 120, a storage unit 130, and an output unit 140. The document classification device 100 is a computer system having a function of determining, when new document data is input, whether the new document data is noise corresponding to a document desired by a user.

The acquisition unit 110 has a function of acquiring information indicating a patent document as new document data that the document classification device 100 classifies. The information indicating the patent document may be any information indicating the patent document, and may be identification information indicating the patent document or the document itself. The information indicating the patent document is provided with the information indicating the patent classification as the second feature information indicating the technical information of the patent document. The acquisition unit 110 is, for example, a communication interface that acquires other unclassified document data from an external device (not shown).

The control unit 120 is a processor having a function of controlling each unit of the document classification device 100 by executing various programs stored in the storage unit 130. The control unit 120 has a function as an extraction unit 121 and a determination unit 122. The control unit 120 determines whether the IPC assigned to each document is “noise” or “corresponding” in order to determine whether the document is a document desired by the user as a document searched according to the search formula. It has a function of generating IPC relevant information as prior information for determining whether to perform. Further, the control unit 120 uses the functions of the extraction unit 121 and the determination unit 122 to determine whether or not the patent document newly searched by the search formula and to which the classification of whether it is the noise or not is not a document desired by the user. It also has a function to judge whether or not.

The extraction unit 121 extracts a document used for determining whether or not a new document is noise from the past document list which is a past classified document list. The extraction unit 121 extracts a predetermined number of documents from the top of the past document list sorted in descending order of the degree of IPC matching with a new document.

The determination unit 122 determines whether or not the new document is a desired document for the user based on the classification (“noise” or “corresponding”) as relevant information given to the document extracted by the extraction unit 121. That is, it is determined whether it is “noise” or “corresponding”. The determination unit 122 sets the classification that occupies the majority of the documents extracted by the extraction unit 121 as the new document classification.

Details of the process of determining whether or not the new document by the control unit 120 is a document desired by the user and the process of generating the IPC relevant information will be described later.

The storage unit 130 is a recording medium having a function of storing various data and programs required for the operation of the document classification device 100. The storage unit 130 is realized by, for example, an HDD, SSD, flash memory, or the like, but is not limited to these. The storage unit 130 corresponds to, for example, a program for the control unit 120 to generate advance information as to whether each IPC is noise or corresponds, or whether the new document is noise when a new document is input. A program for the control unit 120 to determine whether or not is stored. Further, the storage unit 130 is a list of past patent documents, and a past document list 300 in which relevant information indicating whether or not each document corresponds to a document desired by the user is associated with the acquisition unit 110. The new document list 350, which is a list of new documents to be acquired, and the IPC relevant information 500, which is the prior information generated by the control unit 120, are stored. Further, the matching degree table 400 generated when determining whether the new document is noise or relevant is also stored.

The output unit 140 is a communication interface having a function of outputting information regarding the determination result of a new document of the control unit 120 to an external device. Here, for example, it is assumed that the new document list to which the classification is added is output in the mode as shown in FIG. 3A (at least the mode in which the publication number of a new document and the classification are associated with each other).

The above is the description of the configuration of the document classification device 100.

<Data>
From here, various data used in the document classification device 100 will be described.

FIG. 3A is a data conceptual diagram showing a configuration example of a past document list 300 regarding classified document data stored in the storage unit 130. The past document list 300 is information about documents retrieved by a predetermined search formula in the past, and includes information indicating whether or not each document is a desired document for the user. As shown in FIG. 3A, the past document list 300 is searched for a publication number 301 as the first document information indicating a plurality of document data searched according to a search expression and a patent document corresponding to the document data. As a result, the classification 302, which is information corresponding to the corresponding information indicating whether or not the content desired by the user is described, and one or more technical features assigned to each of the document data. This is information associated with the IPC classification 303 corresponding to the characteristic information.

The publication number 301 is document data retrieved according to a retrieval formula, and is information for uniquely identifying a patent document that is classified document data. Here, the publication number of the patent document to be classified is used, but this is not limited to the publication number, and if identification information that can uniquely identify the document is used, other than the publication number is used. Good.

The classification 302 is information that should be called corresponding information indicating whether or not the corresponding patent document is a desired document for the user. Here, when the corresponding patent document is a desired document for the user, If the document is not a desired document, it is represented by a binary value of "noise".

The IPC classification 303 is information indicating the IPC assigned to the corresponding patent document. The IPC classification 303 is information indicating one or more technical characteristics of the corresponding patent documents, and is information indicating the classification according to the technical content of the patent documents that are internationally unified and used.

FIG. 3B is a data conceptual diagram showing a configuration example of a new document list 350 showing an example of new document data acquired by the acquisition unit 110 of the document classification device 100. Each patent document described in the new document list 350 is a list of new document data to be classified. The new document list 350 corresponds to the publication number 351 corresponding to the information indicating other document data newly searched by the search and the IPC classification 352 corresponding to the second characteristic information indicating one or more technical characteristics. It is the attached information.

The publication number 351 is information for uniquely identifying the document data. Here, the publication number of the patent document to be classified is used, but this is not limited to the publication number, and if identification information that can uniquely identify the document is used, other than the publication number is used. Good.

The IPC classification 352 is information indicating the IPC assigned to the corresponding patent document. The IPC classification 352 is information indicating one or more technical features of the corresponding patent documents, and is information indicating the classification according to the technical contents of the patent documents that are internationally unified and used.

FIG. 4 shows a state in which one document included in the new document list 300 and each document included in the past document list 350 are associated with each other in the degree of coincidence in technical classification, and the documents are rearranged in descending order of the degree of coincidence. FIG. 6 is a data conceptual diagram showing a configuration example of a matching degree table 400. The matching degree table 400 is information generated by the document classification device 100 in the process of determining whether new document data is “noise” or “corresponding”.

The publication number 401 is information for uniquely identifying a patent document that is classified document data. Here, the publication number of the patent document to be classified is used, but this is not limited to the publication number, and if identification information that can uniquely identify the document is used, other than the publication number is used. Good.

The classification 402 is information that should be called corresponding information indicating whether or not the corresponding patent document is a desired document for the user. Here, when the corresponding patent document is a desired document for the user, If the document is not a desired document, it is represented by a binary value of "noise".

The IPC classification 403 is information indicating the IPC assigned to the corresponding patent document. The IPC classification 403 is information indicating one or more technical features of the corresponding patent documents, and is information indicating the classification according to the technical contents of the patent documents that are internationally unified and used.

The degree of coincidence 404 is, for one new document included in the new document list 350, the IPC classification 352 assigned to the new document and the IPC classification 303 associated with each document included in the past document list 300. Is information indicating the degree of coincidence.

The matching degree table 400 is generated for each document included in the new document list 350. Then, for each document, the determination unit 122 determines whether each document is a document desired by the user or not, using the corresponding matching degree table 400.

FIG. 5 is a data conceptual diagram showing a configuration example of the IPC relevant information 500 showing the probability of becoming noise or being relevant for each IPC classification. As shown in FIG. 5, the IPC applicable information 500 is information in which the IPC classification 501, the number of publications 502, the number of noise cases 503, the noise ratio 504, and the noise determination 505 are associated with each other.

The IPC classification 501 is information indicating the technical characteristics of the patent documents, and is information indicating the classification according to the technical contents of the patent documents that are internationally unified and used. The IPC classification 501 is an extraction of IPCs assigned to past documents included in the past document list 300.

The number of publications 502 is information indicating the total number of publications to which the corresponding IPC classification 501 is attached in the past publication list 300.

The noise number 503 is information indicating the total number of documents that the user has determined to be “noise” among the past documents to which the corresponding IPC classification 501 is assigned.

The noise ratio 504 is information that indicates a probability that the document is not a document desired by the user when the corresponding IPC classification 501 is added to the document data, and the corresponding noise number 503 is the corresponding publication number 502. The divided value is shown.

The noise determination 505 is a value for determining whether or not the noise is generated when the corresponding IPC classification 501 is given. If the noise determination is “100”, the document classification apparatus 100 can determine “noise”, that is, a document that is not desired by the user. Further, the document classification apparatus 100 can determine that the document is “corresponding”, that is, the document desired by the user if the noise determination is “0”. In the present embodiment, the noise determination 505 is 100 for the IPC classification with the noise ratio 504 of 95 or more, and the noise determination 505 is 0 for the IPC classification with the noise ratio 504 of 10 or less. The thresholds of 95 and 10 are values determined by the document classification device 100, and the settings may be changed as appropriate. When changing the setting, it can be changed using an input device or the like connected to the document classification device 100.

The IPC applicable information 500 can be used by the document classification device 100 to identify a document that can be noise or applicable with high probability for a specific IPC classification. That is, for example, the document classification apparatus 100 can specify a document to which an IPC classification with a noise determination of 100 is added as a document that is not desired by the user. On the contrary, the document to which the IPC classification with the noise determination of 0 is added can be specified as the document desired by the user. Each value shown in FIG. 5 is an example.

<Operation>
From here, the operation related to the calculation of the distortion amount by the document classification device 100 will be described. The flowcharts shown in FIGS. 6 to 7 are flowcharts showing a process for advance preparation before the document classification device 100 classifies a new document. The process is a process executed by the control unit 120 of the document classification device 100. This process is a process for generating the hit probability information 500 shown in FIG. The details will be described below.

(Step S601)
In step S601, the control unit 120 of the document classification device 100 sets the variable i used in the process to 1. The variable i is a variable for determining a document to be processed for each document included in the past document list 300. After setting the variable i to 1, the process proceeds to step S602.

(Step S602)
In step S602, the control unit 120 determines whether or not all the documents included in the past document list 300 have been processed. The determination can be made based on whether or not the number of variables i matches the total number of the past document list 300. In the determination, if all documents have not been processed (NO), the process proceeds to step S603, and if all documents have been processed (YES), the process proceeds to step S609. Transition.

(Step S603)
In step S603, the control unit 120 determines whether or not the category of the publication on the i-th row of the past document list 300 is “noise” by referring to the corresponding category 302 of the past document list 300. If the publication in the i-th row is classified as "noise" (YES), the process proceeds to step S604. If it is not "noise", that is, "corresponding" (NO), step The processing moves to S605.

(Step S604)
In step S604, the control unit 120 sets the count setting value C to 1, and the process proceeds to step S606.

(Step S605)
In step S605, the control unit 120 sets the count setting value C to 0, and proceeds to step S607.

(Step S606)
In step S606, the count setting value C calculated in step S604 or step S605 is added to the noise count of each IPC shown in the IPC classification 303 corresponding to the i-th document in the past document list 300. Here, the noise count is a value that serves as an index for determining whether each IPC is noise. Then, it transfers to step S607.

(Step S607)
In step S607, the control unit 120 adds 1 to the total count value indicating the total number for each IPC shown in the IPC classification corresponding to the i-th document in the past document list 300. Then, it transfers to step S608.

(Step S608)
In step S608, the control unit 120 sets the value obtained by adding 1 to the variable i as the value of the next i, and returns to the process of step S602.

(Step S609)
In step S609, the control unit 120 sets the threshold value T to 2.5% of the total number of documents, which is the total number of documents described in the past document list 300, and proceeds to the process of step S610. Here, the threshold value T is set for each IPC.

(Step S610)
In step S610, the control unit 120 determines whether the threshold T calculated in step S609 exceeds 50. If the threshold T exceeds 50 (YES), the process proceeds to step S611, and if not (NO), the threshold is left as it is and the process proceeds to step S701 of FIG. To do.

(Step S611)
In step S611, the control unit 120 resets the threshold value T to 50, and proceeds to the process of step S701 in FIG.

(Step S701)
In step S701 shown in FIG. 7, the control unit 120 sets the variable j to 1 and shifts to the processing of step S702. The variable j is a variable for specifying the IPC to be processed for each IPC.

(Step S702)
In step S702, the control unit 120 determines whether or not the variable j is equal to the total number of IPCs to be processed plus one. When the variable j is the same as the number obtained by adding 1 to the total number of IPCs to be processed (YES), the process is terminated, and when the number is not the same, the process proceeds to the process of step S703.

(Step S703)
In step S703, the control unit 120 determines whether or not the total number of cases is less than the threshold value T for each IPC. If the total number is less than the threshold T (YES), the process proceeds to step S709, and if it is not less than the threshold T (NO), the process proceeds to step S704.

(Step S704)
In step S704, the control unit 120 calculates the noise rate of each IPC as a value obtained by dividing the noise count value of each IPC by the total count value of the IPC. The noise count value is a value calculated by the process of step S606 by repeating the processes of steps S602 to S608 of FIG. The total count value of IPC is a value calculated by the process of step S607 by repeating the processes of steps S602 to S608. After calculating the noise rate, the process proceeds to step S705.

(Step S705)
In step S705, the control unit 120 determines whether or not the noise rate of each IPC is less than 10%. When it is less than 10% (YES), the process proceeds to step S706, and when it is not less than 10%, that is, when it is 10% or more (NO), the process proceeds to step S708.

(Step S706)
In step S706, the control unit 120 sets the noise determination of the IPC whose IPC noise rate is less than 10% to 0%. Then, the process proceeds to step S709.

(Step S707)
In step S707, the control unit 120 determines whether or not the noise rate is 95% or more for each IPC whose IPC noise rate is not less than 10%. When the noise rate is 95% or more (YES), the process proceeds to step S709, and when the noise rate is not 95% or more (NO), the noise rate is the value calculated in step S704. The process moves to step S709.

(Step S708)
In step S708, the control unit 120 sets the noise determination of the IPC whose IPC noise rate is 95% or more to 100%. Then, the process proceeds to step S709.

(Step S709)
In step S709, the control unit 120 sets the value obtained by adding 1 to j as a new j, and returns to the process of step S702.

By executing the above processing, the control unit 120 calculates the noise ratio for each IPC, generates the IPC applicable information 500 shown in FIG. 5, and stores it in the storage unit 130.

Next, the operation when the document classification device 100 determines whether the publication is “noise” or “corresponding” when new document data (publication) is actually input will be described. The flowcharts shown in FIGS. 8 to 10 correspond to the processing. This process is a process executed by the extraction unit 121 and the determination unit 122 after the acquisition unit 110 acquires the new document list 350 that is a new set of documents. The details will be described below.

(Step S801)
In step S801, the determination unit 122 sets a variable 1 for distinguishing an undetermined document to 1. Then, the process proceeds to step S802.

(Step S802)
In step S802, the determination unit 122 determines whether or not an undetermined document remains. The determination is performed based on whether the number of documents included in the new document list 350 and the variable l match. If there is any undetermined document (YES), the process proceeds to step S803, and if no document remains (NO), the process ends.

(Step S803)
In step S803, the determination unit 122 extracts the IPC of the 1st publication. Here, it is extracted from the IPC classification 352 of the new document list 350. The extracted IPC is managed individually. After extracting the IPC, the process proceeds to step S804.

(Step S804)
In step S804, the determination unit 122 sets 1 to the variable m used to identify the IPC that has not been processed. Here, the maximum value of the variable m corresponds to the total number of IPCs extracted in step S803. Then, the process proceeds to step S805.

(Step S805)
In step S805, the determination unit 122 determines whether or not the determination is for the last IPC, that is, whether m is the total number of m plus 1. If m is equal to the total number of IPCs plus 1 (YES), the process proceeds to step S806, and if not (NO), the process proceeds to step S812.

(Step S806)
In step S806, the determination unit 122 determines whether or not the number of matching IPCs with the search expression IPC is m. If so (YES), the process proceeds to step S807, and if not (NO), the process proceeds to step S901 in FIG.

(Step S807)
In step S807, the determination unit 122 sets the noise determination of the target IPC to 0%, and proceeds to the process of step S811.

(Step S808)
In step S808, the determination unit 122 determines whether or not the noise determination is 0. If the noise determination is 0 (YES), the process proceeds to step S809, and if it is not 0 (NO), the process proceeds to step S810. Incidentally. Here, the determination is made based on whether or not the noise determination is 0. However, this may be determined based on whether or not the noise determination is 100, and when the noise determination is 100, The process moves to step S810, and if not 100, the process moves to step S809.

(Step S809)
In step S809, the determination unit 122 determines that Pl, that is, the l-th document is “corresponding”, that is, the document desired by the user, and stores it in association with the l-th document. Then, the process proceeds to step S811.

(Step S810)
In step S810, the determination unit 122 determines that Pl, that is, the l-th document is “noise”, that is, not the document desired by the user, and stores it in association with the l-th document. Then, the process proceeds to step S811.

(Step S811)
In step S811, the determination unit 122 sets the value obtained by adding 1 to the variable l as a new variable l, and returns to the process of step S802.

(Step S812)
In step S812, the determination unit 122 determines whether the noise rate of the processing target IPC is set to 0% or 100% in the IPC applicable information 500. If the noise rate is set to 0% or 100% (YES), the process proceeds to step S813, and if it is not set (NO), the process proceeds to step S814.

(Step S813)
In step S813, the determination unit 122 sets the noise determination of the new document to the noise determination value of the target IPC (that is, either 0% or 100%), and proceeds to the processing of step S808. In the processing, when a new document has an IPC classification of 100% noise or an IPC classification of 100% in the IPC relevant information 500, the classification shown in the IPC relevant information 500 is applied to the new document as it is. To do.

(Step S814)
In step S814, the determination unit 122 determines whether the IPC to be processed matches the IPC used as the search expression. If they match (YES), the process proceeds to step S815, and if they do not match (NO), the process proceeds to step S816.

(Step S815)
In step S815, the determination unit 122 adds 1 to the IPC number count that matches the search expression IPC. Then, the process proceeds to step S816.

(Step S816)
In step S816, the determination unit 122 sets the value obtained by adding 1 to the variable m as a new m, and returns to the process of step S805.

(Step S901)
In step S901, the determination unit 122 sets the variable k to 1. The variable k is a variable used to identify past documents in the past document list 300 to be processed. After setting the variable k to 1, the process proceeds to step S902.

(Step S902)
In step S902, the determination unit 122 determines whether or not the number of the past documents list of the past documents list 300 is the number of the past documents list of the past documents list 300. It is determined by whether or not the value matches If the document to be processed is the last document in the past documents included in the past document list 300 (YES), the process proceeds to step S1001 in FIG. 10, and if not (NO), step Then, the process proceeds to S903.

(Step S903)
In step S903, the determination unit 122 extracts the IPC of the kth publication in the past literature list 300. That is, each IPC is extracted from the IPC classification 303 on the k-th row of the past document list 300. Then, the process proceeds to step S904.

(Step S904)
In step S904, each IPC(h) of the h-th publication in the new document list 350 is extracted from the IPC classification 352 of the new document list 350. Then, the process proceeds to step S905.

(Step S905)
In step S905, the determination unit 122 sets the variable n to 1. The variable n is a variable for distinguishing the IPC to be processed among the IPCs assigned to the new document to be processed. After setting the variable n to 1, the process proceeds to step S906.

(Step S906)
In step S906, the determination unit 122 determines whether n is the last, that is, whether or not processing has been performed for all IPCs assigned to the new document. If so (YES), the process proceeds to step S906, and if not (NO), the process proceeds to step S907.

(Step S907)
In step S907, the determination unit 122 determines whether IPC(h)n matches the IPC of the search expression. When they match (YES), the process proceeds to step S908, and when they do not match (NO), the process proceeds to step S909.

(Step S908)
In step S908, the determination unit 122 subtracts 1 from the target IPC count, and shifts to the processing in step S911.

(Step S909)
In step S909, the determination unit 122 determines whether IPC(h)n matches IPC(k). That is, of the IPCs assigned to the h-th new document in the new document list 350, whether the n-th IPC matches any of the IPCs assigned to the k-th past documents in the past document list 300. Determine whether. If they match, the process proceeds to step S910 (YES), and if they do not match (NO), the process proceeds to step S911.

(Step S910)
In step S910, the determination unit 122 increments the IPC coincidence count by 1, and then proceeds to step S911.

(Step S911)
In step S911, the determination unit 122 sets the value obtained by adding 1 to k as a new k, and returns to the process of step S902.

(Step S912)
In step S912, the determination unit 122 has counted the concordance rate between the IPC of the h-th new document in the new document list 350 and the IPC assigned to the k-th past document in the past document list 300 by then. It is calculated by dividing the IPC match rate count by the target IPC count number. Then, the process proceeds to step S913.

(Step S913)
In step S913, the determination unit 122 sets the value obtained by adding 1 to the variable k as a new k, and returns to the process of step S902.

(Step S1001)
In step S1001, the determination unit 122 arranges the concordance rate for each document in descending order with respect to the IPC assigned to the new document in the new document list 300 and the IPC assigned to each past document in the past document list 350. Change. Then, the process proceeds to step S1002.

(Step S1002)
In step S1002, the extraction unit 121 sets the variable q to 1. The variable q is a variable for specifying the number for extracting past documents from the one having a high degree of coincidence with the IPC. After setting the variable q to 1, the process proceeds to step S1003.

(Step S1003)
In step S1003, the extraction unit 121 determines whether q is 8. If q is 8 (YES), the process proceeds to step S1004, and if not (NO), the process proceeds to step S1009.

(Step S1004)
In step S1004, the determination unit 122 divides the noise count by the comparative publication number count for the index t for determining whether or not the target newly searched patent document is a document desired by the user. Calculate with. The noise count is the number calculated in step S1011 and indicates the number of documents that are noise in the past documents extracted by a predetermined number from the top of the documents that have a high degree of agreement in the patent classification. The comparative publication number count is the number counted in step S1012, and matches the maximum number of q. That is, the count of comparative publications means the number of publications to be extracted. When t is calculated, the process proceeds to step S1005.

(Step S1005)
In step S1005, the determination unit 122 determines whether t calculated in step S1004 exceeds a predetermined threshold value α. If t exceeds the threshold value α (YES), the process proceeds to step S1006, and if t does not exceed (NO), the process proceeds to step S1007.

(Step S1006)
In step S1006, the determination unit 122 adds information indicating that the corresponding new publication corresponds to the document desired by the user (classifies it as applicable). Then, the process proceeds to step S1008.

(Step S1007)
In step S1007, the determination unit 122 assigns information indicating that the corresponding new publication is noise, which is not a document desired by the user (classifies it as noise). Then, the process proceeds to step S1008.

(Step S1008)
In step S1008, the determination unit 122 sets the value obtained by adding 1 to the variable l as a new l, and proceeds to the process of step S802 in FIG.

(Step S1009)
In step S1009, the determination unit 122 determines whether the number of documents to be processed has reached a value obtained by adding one to the total number of the past document list 300. The determination is a treatment for the case where the past document list 300 does not include q documents. If the number of documents to be processed has reached the value obtained by adding one to the total number of the past document list 300 (YES), the process proceeds to step S1004, and if not (NO), the process of step S1010. Move to.

(Step S1010)
In step S1010, the determination unit 122 determines whether or not the classification 302 of the q-th publication in the past document list is “noise”. If it is determined that it is noise (YES), the process proceeds to step S1011. If it is determined that it is not noise (NO), the process proceeds to step S1012.

(Step S1011)
In step S1011, the determination unit 122 adds 1 to the noise count, and the process proceeds to step S1012.

(Step S1012)
In step S1012, the determination unit 122 adds 1 to the comparative publication number count, and the process proceeds to step S1013.

(Step S1013)
In step S1013, the determination unit 122 sets the value obtained by adding 1 to the variable q as a new q, and proceeds to the process of step S1003.

By executing the processes shown in FIGS. 8 to 10, for all the new documents included in the new document list 350, the document classification device 100 can determine whether or not each new document is noise. it can.

The above is the description of the operation of the document classification device 100.

<Summary>
The document classification device according to the above-mentioned embodiment, based on the patent classification originally assigned to the patent publication, whether the document obtained by the search formula in advance is a desired one, "noise", "corresponding Classification information is added. Then, when a new patent publication is input, a predetermined number of documents are extracted based on the degree of coincidence between the patent classification assigned to the new patent publication and the patent classification of the classified patent publication. Then, depending on which of “noise” and “corresponding” is assigned to the extracted document, whether the new patent publication is “noise” or “corresponding” is determined by the content of the patent publication. Can be classified without scrutiny. Then, the user does not need to check the content of the document determined to be noise even if the document is searched according to the search formula set by the user, so the time required for screening the document is shortened. can do. Further, as the document classification device, it is not necessary to scrutinize the gazette (it is not necessary to perform morphological analysis or see the matching rate of a huge number of words extracted by morphological analysis). The processing load of the processor can be reduced as compared with the classifying devices shown in 1 to 3.

<Supplement>
It goes without saying that the document classification device according to the above embodiment is not limited to the above embodiment and may be realized by other methods. Various modifications will be described below.

(1) Although not particularly described in the above embodiment, it is desirable that the number of documents q extracted by the extraction unit 121 is an odd number. This is because it is always possible to specify “noise” or “corresponding” by setting an odd number. The variable q may be calculated using the so-called k-nearest neighbor method.

In addition, when q is set to an even number, the number of “noise” and the number of “corresponding” may be the same. Therefore, the document classification device may classify the documents using the following method. That is, the basic value of the "noise" document is "-1", and the basic value of "corresponding" is "+1". Then, a value obtained by multiplying the basic value by the degree of coincidence as a weight value is set as a classification value of noise or relevant in the document. Then, the determination unit 122 may classify the classification values of the documents extracted by the extraction unit 121, classify as “corresponding” if the value is positive, and classify as “noise” if the value is negative. Good. In the case of the method, although the processing load related to the determination processing of “noise” or “corresponding” is larger than the processing load of the processing described in the above-described embodiment, it is more accurate whether the processing is “noise” or “corresponding”. Can make decisions. That is, the document classification device 100 may perform the classification after performing the correction by weighting. Note that, here, the degree of coincidence itself is used as the weight value, but this is not the limitation, and any value may be used as the weight value.

(2) In the above-mentioned embodiment, the classification of the majority of the judgments as to whether it is noise or not is the classification of the new document, but this is not the limitation. For example, among the q documents extracted by the extraction unit 121, if a predetermined number or more of the documents are classified as “noise”, the new document may be determined to be “noise”. For example, the number of extracted documents may be set to 10, and if eight or more of them are classified as “noise”, a new document may be classified as “noise”.

(3) In the above-described embodiment, in determining whether the patent classification, which is a technical feature, is noise or the corresponding, the classification having a noise rate of less than 10% is set to 0% noise determination, and the IPC noise is determined. The classification with the rate of 95% or more was set as the noise determination of 100%. Here, it can be said that the threshold of 10% is the second threshold for determining whether or not the document corresponds to the document desired by the user when the corresponding classification is given. That is, it can be said that the determination in step S705 is a determination as to whether the corresponding rate is 90% or more. The same applies to the first threshold used for the determination in step S707.

That is, the document classification apparatus 100 makes a determination based on whether or not the applicable rate indicating whether or not the patent classification given to the document is applicable is 90% or more, which is the first threshold, and the non-applicable rate is It can be understood that the determination is made based on whether the threshold value is 95% or more, which is two threshold values. Here, by providing a difference between the first threshold value and the second threshold value, it is possible to ensure that the classification can be classified as either noise or relevant. Further, the first threshold value and the second threshold value may be changed depending on whether to give priority to the determination that the classification is noise or to give the determination that the classification is applicable. Therefore, the document classification device 100 may include a setting unit for setting the first threshold and the second threshold. The input to the setting unit may be set to an appropriate value by the document classification device 100 through learning, or may be set by the user of the document classification device 100. Note that the threshold percentages used for these determinations are not limited to the numerical values shown in the above embodiments, and the set values can be changed by the operator of the document classification apparatus 100 as appropriate.

(4) In the above-described embodiment, a document to which the IPC noise rate is assigned 100% is classified as noise, and a document to which the IPC noise rate is 0% is assigned. I have decided. However, in some cases, there may be a document to which a noise rate of 100% and a noise rate of 0% are assigned. In such a case, the document classification device 100 may determine that the document is “noise” or may be “corresponding” according to a predetermined standard determined by the user in advance. Good. For example, when “noise” is set as a priority, it may be determined as “noise”, and when “corresponding” is set as a priority, it may be determined as “corresponding”.

(5) In the above embodiment, as a method for the document classification device to classify new document data, the processor functioning as each functional unit constituting the document classification device 100 executes the document classification program or It is intended to classify various document data, but this is realized by a logic circuit (hardware) formed in an integrated circuit (IC (Integrated Circuit) chip, LSI (Large Scale Integration)) or a dedicated circuit in the device. Good. Further, these circuits may be implemented by one or a plurality of integrated circuits, and the functions of the plurality of functional units described in the above embodiment may be implemented by a single integrated circuit. The LSI may be referred to as VLSI, super LSI, ultra LSI, or the like depending on the degree of integration. That is, as shown in FIG. 11, each functional unit configuring the document classification device 100 may be realized by a physical circuit. As shown in FIG. 11, the document classification device 100 includes a storage circuit 130a, an acquisition circuit 110a, an extraction circuit 121a, a determination circuit 122a, and an output circuit 140a, and each circuit has the same function as described above. It has the same function as the section.

Further, the document classification program may be recorded in a processor-readable recording medium, and the recording medium may be a “non-transitory tangible medium”, for example, a tape, a disk, a card, a semiconductor memory, a programmable logic. A circuit or the like can be used. Further, the document classification program may be supplied to the processor via any transmission medium (communication network, broadcast wave, etc.) capable of transmitting the document classification program. The present invention can also be realized in the form of a data signal embedded in a carrier wave, in which the document classification program is realized by electronic transmission.

The document classification program can be implemented using, for example, a script language such as ActionScript or JavaScript (registered trademark), an object-oriented programming language such as Objective-C or Java (registered trademark), or a markup language such as HTML5. ..

(6) The configurations shown in the above-described embodiment and each supplement may be appropriately combined.

100 document classification device 110 acquisition unit 121 extraction unit 122 determination unit 140 output unit

Claims (9)

The first document information indicating a plurality of document data searched according to the search formula, the first feature information indicating one or more technical features of the document data attached to the document data, and the document data are A storage unit that stores corresponding information indicating whether or not the document is a desired document for the user,
An acquisition unit that acquires second document information indicating another document data that is newly searched according to the search formula and that is provided with second characteristic information indicating one or more technical characteristics;
An extraction unit that extracts a predetermined number of document data from the plurality of document data based on the degree of coincidence between the second characteristic information and the first characteristic information,
A determination unit that determines whether or not the other document data is a document desired by the user, based on the corresponding information associated with each of the predetermined number of document data extracted by the extraction unit,
A document classification device, comprising: an output unit that outputs a determination result of the determination unit. The extraction unit extracts, from the plurality of document data, a predetermined number from the one having a high degree of coincidence between the second characteristic information and the first characteristic information,
When the corresponding information associated with the document extracted by the extraction unit has more than a threshold value indicating that the document is desired by the user, the determination unit determines that the other document data is desired by the user. The document is determined to be a document, and when there are more than a threshold value indicating that the document is not a document desired by the user, it is determined that the other document data is not a document desired by the user. Document classification device described. The determination unit performs weighting on the relevant information associated with the documents extracted by the extraction unit according to the degree of matching, and based on the relevant information after performing the weighting, the other document. The document classification device according to claim 1 or 2, wherein it is determined whether or not the data is a desired document for the user. The determination unit is associated with the first feature information of the document data that is associated with the first feature information and indicates that the corresponding information is a document that the user does not desire. The non-correspondence rate indicating the ratio to the entire document data that the user does not want other document data having the second feature information that matches the first feature information that exceeds the first threshold value. The document classification device according to claim 1, wherein the document classification device is a document classification device. The determination unit is associated with the first feature information of the document data that is the document data associated with the first feature information and indicates that the corresponding information is a document desired by the user. That the document is desired by the user for other document data having the second feature information that matches the first feature information that exceeds the second threshold. The document classification device according to claim 4, wherein the document classification device is a document classification device. The document classification device according to claim 5, wherein the first threshold value is larger than the second threshold value. When the feature information is used in the search formula, the determination unit makes the determination based on the first feature information excluding the feature information and the second feature information. The document classification device according to any one of claims 1 to 6. The first document information indicating a plurality of document data searched according to the search formula, the first feature information indicating one or more technical features of the document data attached to the document data, and the document data are A storage step of storing corresponding information indicating whether or not the document is a desired document for the user,
An acquisition step of acquiring second document information indicating another document data newly searched in accordance with the search formula and provided with second characteristic information indicating one or more technical characteristics;
An extraction step of extracting a predetermined number of document data from the plurality of document data based on the degree of coincidence between the second feature information and the first feature information;
A determination step of determining whether or not the other document data is a document desired by the user, based on the corresponding information associated with each of the predetermined number of document data extracted in the extraction step,
An output step of outputting the result of the determination in the determination step. On the computer,
The first document information indicating a plurality of document data searched according to the search formula, the first feature information indicating one or more technical features of the document data attached to the document data, and the document data are A storage function of storing corresponding information indicating whether or not the document is a desired document for the user,
An acquisition function for acquiring second document information indicating another document data newly searched in accordance with the search formula and provided with second characteristic information indicating one or more technical characteristics;
An extraction function of extracting a predetermined number of document data from the plurality of document data based on the degree of coincidence between the second feature information and the first feature information,
A determination function of determining whether or not the other document data is a document desired by the user, based on the corresponding information associated with each of the predetermined number of document data extracted by the extraction function,
A document classification program that realizes an output function of outputting a determination result of the determination function.

Images