JP5310196B2 - Classification system revision support program, classification system revision support device, and classification system revision support method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To support revision of an efficient classification system while maintaining overall consistency. <P>SOLUTION: A common viewpoint field retrieval means 2 extracts at least two fields each provided with a category in a common viewpoint in reference to a classification system information storage means 1. A hypothesis definition means 5 compares the categories provided in the common viewpoint of the extracted fields 3, 4, sets the category present only in a part of the field of the extracted fields as a category to be added, and defines a hypothesis that the category is added to another field. A hypothesis verification means 7 detects additional category matching information belonging to the category to be added when the category to be added is added according to the hypothesis, and decides whether the hypothesis is proper or not based on the amount of the detected additional category matching information. The hypothesis verification means 7 outputs the hypothesis decided that it is proper as a revision idea 8. <P>COPYRIGHT: (C)2011,JPO&amp;INPIT

Description

  The present invention relates to a classification system revision support program, a classification system revision support device, and a classification system revision support method that support revision work of a classification system.

  With the development of information communication technology, a huge amount of document information can be easily accessed using a computer. There is a technique for classifying document information in advance in order to easily find a document required by a user from such a large amount of document information. If the document information is classified in advance under an accurate classification system, the user of the document information can find out a necessary document relatively easily from the types of document information according to the purpose. As an example of effectively utilizing such classification of document information, there is a classification system of patent documents (open patent gazette, patent gazette, etc.).

  In many countries, the International Patent Classification (IPC) is used as a classification system for patent documents. Classification information by IPC is given to patent documents issued in each country. In addition to IPC, a plurality of classification information such as FI and F-term are given to patent documents issued in Japan. FI is a further development of IPC. F-term is a search code system developed by the Japan Patent Office. Classification systems such as IPC, FI, and F-term are utilized in prior art searches for patent examinations within the Japan Patent Office and various technical searches in companies.

  Among these patent classification systems, especially in the F term, various viewpoints such as fine classification of technical elements unique to each field are set. Further, in the F term, a number indicating a category for classification based on a classification viewpoint is set under the viewpoint. A viewpoint and a category set based on the viewpoint are set for each field.

  By the way, in the classification based on F-terms, categories may be set from the same viewpoint for different fields. At this time, although the category is subdivided from the same viewpoint in the field A and the field B different from A, it is defined only under the field A and defined under the field B. There may be no categories. If the consistency and consistency of the classification system is not maintained in this way, the convenience of the user who performs the search and analysis using the F term will be reduced.

  As a technique for supporting the review of the classification system so as to improve the convenience for the user, for example, there is a technique for evaluating the classification according to the usage status of the user. In this technology, information that is frequently referenced by users is classified into information that can be seen by even more users, and access history of users is recorded, and each information is recorded based on this recorded access history. There is a technique for calculating an evaluation value of classification.

  In addition, with reference to information that has already been classified into groups, there is also a technology that identifies information that is missing in the newly created classification structure and presents the missing information to the user as a candidate for addition to the classification structure. is there.

JP 2006-92214 A JP 2006-39810 A JP 2006-85571 A

  However, in the conventional classification system review technology, it is not possible to perform a review that improves the consistency of categories among a plurality of fields in which a common category can be set from the same viewpoint. It is necessary for search and analysis across multiple fields if there are differences in the categories that are actually set in each field, even if the same category can be set in multiple fields if they are classified from the same viewpoint. More time is required. For example, when searching across multiple fields, it is necessary to set individual search conditions for each field. Moreover, if the search conditions are different for each field, when the search results are statistically analyzed, the search results for each field cannot be compared equally.

  If the viewpoints are the same, the consistency as the classification system can be maintained if the categories for classification from the viewpoints are the same in all fields. However, even if the viewpoints are the same, some categories are inappropriate depending on the field. For this reason, if categories for classification from a specific point of view are made common in all fields, many categories that are rarely used depending on the field are generated. If the number of categories in the classification system is increased unnecessarily, it becomes difficult for the user to find an appropriate category when searching for information, and convenience is reduced.

  Such a problem occurs not only in the F-term used for classification of patent documents, but also in a classification system when a large amount of information is classified. For example, Internet search engines include directory search engines that classify accessible documents by category. Even when an information search service is performed using such a directory-type search engine, it is important that the categories in each field have uniformity in order to improve user convenience.

  The present invention has been made in view of such points, and a classification system revision support program and classification system revision that can support revision of a classification system so as to be a wasteful classification system while maintaining overall consistency. An object is to provide a support device and a classification system revision support method.

  In order to solve the above-described problems, a classification system revision support program that causes a computer to execute the following processing is provided. Based on this classification system revision support program, a computer stores classification system information indicating a classification system in which a viewpoint of classification is defined in each of a plurality of fields and a category for classification according to the defined viewpoint is provided. By referring to the system information storage means, at least two fields in which categories are provided from a common viewpoint are extracted. Next, the computer compares the categories provided from the common viewpoints of the extracted fields. Next, the computer defines a hypothesis that a category that exists only in some of the extracted fields is set as an addition target category and the additional target category is added to another field. Next, the computer refers to a database in which information to be classified by the classification system is stored, and detects additional category conforming information that belongs to the additional target category when the additional target category is added according to the hypothesis. Furthermore, the computer determines the suitability of the hypothesis based on the detected amount of additional category conformance information. Then, the computer outputs the hypothesis determined to be appropriate as a revised proposal.

  Of the proposals for unifying classification categories from a common viewpoint in different fields, only the revision proposal for adding a useful category can be presented to the user.

It is a figure which shows the outline | summary of embodiment. It is a figure which shows the example of F term regarding a rice transplanter. It is a figure which shows the example of F term regarding a light source. It is a figure which shows the system configuration example of this Embodiment. It is a figure which shows the hardware structural example of the classification system revision assistance apparatus used for this Embodiment. It is a block diagram which shows the function of a classification system revision assistance apparatus. It is a figure which shows the example of a data structure of F term definition memory | storage part. It is a flowchart which shows the procedure of classification system revision proposal creation processing. It is a figure which shows the example of the consistency examination field pair discovered by the field discovery process containing the same viewpoint. It is a figure which shows a hypothesis setting example. It is a figure which shows the verification example of a hypothesis. It is a block diagram which shows the internal function of a viewpoint common field search part. It is the first half of the flowchart which shows the procedure of a viewpoint common field search process. It is a figure which shows the example of the correction candidate table in a correction candidate memory | storage part. It is the second half of the flowchart which shows the procedure of a viewpoint common field search process. It is a figure which shows the example of the comparison table in a comparison result memory | storage part. It is a figure which shows the comparison table after determination result writing. It is a figure which shows the correction candidate table after priority setting. It is a block diagram which shows the internal function of a hypothesis setting part. It is a flowchart which shows the procedure of a hypothesis setting process. It is a figure which shows the example of the totaled comparison table. It is a figure which shows the example of a data structure of the hypothesis table in a hypothesis memory | storage part. It is a block diagram which shows the internal function of a hypothesis verification part. It is a flowchart which shows the procedure of hypothesis verification processing. It is a flowchart which shows the procedure of the feature keyword extraction process of an additional category. It is a figure which shows the classification | category state of the patent document in a reference | standard field | area. It is a figure which shows the example of a data structure of a characteristic keyword table. It is a figure which shows the example of the hypothesis table after completion of hypothesis verification. It is a figure which shows the example of a data structure of the revision plan memory | storage part. It is a block diagram which shows the internal function of the hypothesis verification part in 2nd Embodiment. It is a flowchart which shows the procedure of the hypothesis verification process using machine learning. It is a figure which shows the example of the classification system of the information regarding a recipe of cooking. It is a figure which shows the classification | category system of the exhibition information regarding an auction via a network.

Hereinafter, the present embodiment will be described with reference to the drawings.
FIG. 1 is a diagram showing an outline of the embodiment. The classification system information storage means 1 stores classification system information in advance. The classification system information shows a classification system in which a viewpoint of classification is defined for each of a plurality of fields, and a category for classification according to the defined viewpoint is provided.

  The viewpoint common field search means 2 refers to the classification system information storage means 1 and extracts at least two fields in which categories are provided from a common viewpoint. In the example of FIG. 1, two fields 3 and 4 having a common viewpoint of “purpose / function” are extracted.

  The hypothesis definition means 5 compares the categories provided from the common viewpoints of the extracted fields 3 and 4. Then, the hypothesis definition means 5 defines a hypothesis in which a category that exists only in a part of the extracted fields is set as an addition target category and the addition target category is added to another field. In the example of FIG. 1, categories such as “improvement of operability” and “improvement of safety” are provided in both of the two fields 3 and 4, but the category of “improving planting accuracy” It exists only in the field 3 called “grounding part”. Therefore, the hypothesis definition means 5 defines a hypothesis that a category of “improving planting accuracy” is added regarding the classification from the viewpoint of “purpose / function” in the field 4 of “elements of rice transplanter”.

  The hypothesis verification unit 7 refers to the database 6 in which information to be classified by the classification system is stored, and detects additional category conforming information that belongs to the additional target category when the additional target category is added according to the hypothesis. To do. In the example of FIG. 1, information related to “improvement of planting accuracy” is detected from information stored in the database 6 in the field related to “elements of rice transplanter”. Further, the hypothesis verification means 7 determines the suitability of the hypothesis based on the detected amount of additional category conformance information. For example, the hypothesis verification means 7 determines that the hypothesis is appropriate if the number of detected additional category matching information is a predetermined number or more. Then, the hypothesis verification means 7 outputs the hypothesis determined to be appropriate as a revised plan 8. For example, the hypothesis verification means 7 displays the content of the revised plan 8 on a monitor and presents it to the user.

  In this way, when considering a revision of the classification system, candidates for revision can be determined in consideration of how classification is performed in a plurality of fields having a common viewpoint of classification. Therefore, it is possible to present to the user an amendment that maintains the consistency of the classification system across the cross. Moreover, by verifying the hypothesis, no additional proposals are made for unnecessary categories. As a result, it is possible to present to the user only a revised proposal for adding a useful category among unified proposals for classification categories from a common viewpoint in different fields.

  Such a classification system revision support technology can be used effectively for revision of a patent document classification system. Patent literature is first classified by technical field. Patent documents in the same technical field can be classified from various viewpoints. There is an F-term as such a classification system.

  In the F-term, categories to be set are determined for each technical field depending on technical necessity and differences in development direction. Therefore, if the proposed amendment of the classification system by F-term is presented by the system as shown in Fig. 1, the consistency of the category is maintained in the field where similar classification is possible, and it is necessary in the field to be revised. An amendment can be drafted by evaluating the high nature. In other words, it is possible to create a revision plan without waste while maintaining overall consistency.

Therefore, an embodiment for enhancing the consistency of categories between fields regarding F terms used for classification of patent documents will be described in detail below.
[First Embodiment]
First, the F term will be described in detail. In the F term, categories for classification are defined for each field from a predetermined viewpoint.

  FIG. 2 is a diagram illustrating an example of an F term relating to a rice transplanter. Inventions related to rice transplanters are subdivided into fields such as “elements of rice transplanters” and “grounding parts of rice transplanters”. A category code is assigned to each field. The category code of “rice transplanter element” is “2B062”. The category code of “Resting part of rice transplanter” is “2B063”. The category code can uniquely identify the field to be searched.

  In each field, viewpoints for categorization are defined. In the example of FIG. 2, two F-term lists 41 and 42 are shown. The F-term list 41 indicates categories for classifying patent documents related to the field “elements of rice transplanters” from the viewpoint of “purpose / function”. The F-term list 42 indicates a category for classifying patent documents related to the field “grounding part of rice transplanter” in terms of “purpose / function”. Thus, even patent documents in different fields can be classified from the same viewpoint.

  Categories have a hierarchical structure. The number of “•” on the left of the character string indicates the depth of the hierarchy. If there is only one “•”, it is the category of the first hierarchy. If there are two “•”, it is the category of the second hierarchy. If there are three “•”, it is a category of the third hierarchy. A category in a lower hierarchy is a category included in a category in an upper hierarchy than that category. In the F-term lists 41 and 42, the lower hierarchy categories are listed directly below the corresponding upper hierarchy categories. For example, the category of “improvement of operability” in the F-term list 41 is included in the category of “improvement of performance of transplanter and the like”.

  In addition, a category “others” is provided as a category of classification from each viewpoint. In the “other” category, patent documents not included in the set category are classified.

  In this way, the field of “rice transplanter element” (theme code “2B062”) and “landering part of rice transplanter” (theme code “2B063”) both have the viewpoint of “purpose / function”. . Therefore, compare the categories in the F-term when classifying the “rice transplanter element” and the “rice transplanter grounding part” in terms of “purpose / function”.

  “Performance / Improvement of transplanter etc.” is set as a category in terms of “Purpose / Function” of “Elementary Transplant Elements” and “Grounding Section of Rice Transplanter”, and further under that, “Improved operability” And categories such as “Improving safety” are set.

  Here, the category of “improving planting accuracy” defined in “grounding part of rice transplanter” is not set to “elements of rice transplanter”, the field area is close, and the viewpoint of classification is the same. Even if it exists, it is not consistent as a classification system. In addition, some patent documents classified as “elements of rice transplanter” also use the term “planting accuracy”, and “improvement of planting accuracy” under “elements of rice transplanter”. You may need to set a category.

  In the F-term as shown in FIG. 2, for example, a case is considered in which patent documents for rice transplanters in general are searched with an F-term representing “improving planting accuracy”. In this case, even if there is a patent related to both “elements of rice transplanter” and “grounding part of rice transplanter”, patent documents cannot be obtained as search results in “grounding part of rice transplanter”.

Such category inconsistencies between fields also occur outside of rice transplanters.
FIG. 3 is a diagram illustrating an example of an F term relating to a light source. In the example of FIG. 3, three fields having viewpoints related to “light source” are shown. The first field is “optical recording / reproducing 4 (head itself)” having a viewpoint of “configuration of light source” (theme code “5D119”). The second field is “length measuring device by optical means” having the viewpoint of “light source” (theme code “2F065”). The third field is a “laser beam printer” having a viewpoint of “light source” (theme code “2C362”). FIG. 3 shows F-term lists 43 to 45 classified according to the relevant viewpoints of each field.

  Comparing the categories of the second layer into which the types of light sources in the F-term lists 43 to 45 shown in FIG. 3 are classified, “gas laser” and “semiconductor laser” are common. However, other categories do not match.

  Here, attention is paid to the category “lamp” included in the viewpoint “configuration of the light source” (theme code “5D119”) in the field “optical recording / reproducing 4 (head itself)”. The category “lamp” is not included in the viewpoint “light source” (theme code “2F065”) of the field “optical measuring device by optical means”. However, if patent documents actually classified into the theme code “2F065” are searched with the keyword “lamp”, hundreds of patent documents may be hit. In this case, it can be estimated that it is appropriate to add the category “lamp” to the viewpoint “light source” (theme code “2F065”) of the field “optical measuring device by optical means”.

  The category “lamp” is not included in the viewpoint “light source” (theme code “2C362”) of the field “laser beam printer”. However, it is difficult to assume that lamps are used in place of lasers in the light source technology for laser beam printers. Then, it can be estimated that it is inappropriate to add the category “lamp” to the viewpoint “light source” (theme code “2C362”) in the field “laser beam printer”.

  As described above, there are cases where it is appropriate to unify categories and cases where it is inappropriate to unify categories even if they are classified from a common viewpoint in different fields. Therefore, a classification system revision support device capable of automatically extracting a viewpoint that can unify categories in the F-term and determining whether or not to unify categories is described in detail below.

  FIG. 4 is a diagram illustrating a system configuration example of the present embodiment. The classification system revision support device 100 is connected to the patent information server 20 and the plurality of terminal devices 31 to 33 via the network 10. The patent information server 20 has a patent document database (DB) 21. In the patent document DB 21, various patent documents are classified and stored by F-term. Specifically, an F term indicating a category to which the patent document belongs is assigned to each patent document. One patent document can belong to a plurality of categories. Therefore, a plurality of F terms are given to patent documents belonging to a plurality of categories. The patent information server 20 provides an information search service using F-terms to users who use the terminal devices 31 to 33.

  The classification system revision support device 100 is a computer that supports revision of the classification system using F-terms. More specifically, the classification system revision support apparatus 100 searches for pairs of fields that should maintain consistency by using the definition of the F term. Next, the classification system revision support apparatus 100 makes a hypothesis as to what kind of revision should be performed on the obtained pair. Finally, the classification system revision support apparatus 100 narrows down the revision points by performing verification using the information of the patent document to which the F term is actually assigned.

  For example, when there are two fields A and B in which almost the same viewpoint and category are set, the classification system revision support apparatus 100 can set the category set in the field A also under the field B. Assume that The classification system revision support apparatus 100 determines whether or not the revision should be made by verifying the suitability of the category addition based on this assumption.

  FIG. 5 is a diagram illustrating a hardware configuration example of the classification system revision support device used in the present embodiment. The classification system revision support apparatus 100 is entirely controlled by a CPU (Central Processing Unit) 101. A RAM (Random Access Memory) 102 and a plurality of peripheral devices are connected to the CPU 101 via a bus 108.

  The RAM 102 is used as a main storage device of the classification system revision support device 100. The RAM 102 temporarily stores at least part of an OS (Operating System) program and application programs to be executed by the CPU 101. The RAM 102 stores various data necessary for processing by the CPU 101.

  Peripheral devices connected to the bus 108 include a hard disk drive (HDD) 103, a graphic processing device 104, an input interface 105, an optical drive device 106, and a communication interface 107.

  The HDD 103 magnetically writes and reads data to and from the built-in disk. The HDD 103 is used as a secondary storage device of the classification system revision support device 100. The HDD 103 stores an OS program, application programs, and various data. Note that a semiconductor storage device such as a flash memory can also be used as the secondary storage device.

  A monitor 11 is connected to the graphic processing device 104. The graphic processing device 104 displays an image on the screen of the monitor 11 in accordance with a command from the CPU 101. Examples of the monitor 11 include a display device using a CRT (Cathode Ray Tube) and a liquid crystal display device.

  A keyboard 12 and a mouse 13 are connected to the input interface 105. The input interface 105 transmits a signal sent from the keyboard 12 or the mouse 13 to the CPU 101. The mouse 13 is an example of a pointing device, and other pointing devices can also be used. Examples of other pointing devices include a touch panel, a tablet, a touch pad, and a trackball.

  The optical drive device 106 reads data recorded on the optical disk 14 using laser light or the like. The optical disk 14 is a portable recording medium on which data is recorded so that it can be read by reflection of light. The optical disk 14 includes a DVD (Digital Versatile Disc), a DVD-RAM, a CD-ROM (Compact Disc Read Only Memory), a CD-R (Recordable) / RW (ReWritable), and the like.

  The communication interface 107 is connected to the network 10. The communication interface 107 transmits / receives data to / from other computers such as the patent information server 20 via the network 10.

  With the hardware configuration as described above, the processing functions of the present embodiment can be realized. Although FIG. 5 shows a hardware configuration example of the classification system revision support device 100, the patent information server 20 and the terminal devices 31 to 33 can also be realized by similar hardware.

  FIG. 6 is a block diagram illustrating functions of the classification system revision support device. The classification system revision support apparatus 100 includes an F-term definition storage unit 110, a viewpoint common field search unit 120, a correction candidate storage unit 130, a comparison result storage unit 140, a hypothesis setting unit 150, a hypothesis storage unit 160, a hypothesis verification unit 170, a revision A plan storage unit 180 and a result display unit 190 are provided.

  The F-term definition storage unit 110 has a storage function for storing the definition contents of F-terms for classifying patent documents. For example, a part of the storage area of the HDD 103 is used as the F-term definition storage unit 110.

  The viewpoint common field search unit 120 refers to the F-term definition storage unit 110 and searches for two fields having commonality in terms of categorization of each field. Then, the common viewpoint search unit 120 sets the two detected fields as a pair of fields whose consistency should be examined (consistency examination field pair). At this time, the viewpoint common field search unit 120 stores the correction candidate information in which one of the consistency examination field pairs is a correction candidate in the correction candidate storage unit 130. Further, the common viewpoint search unit 120 compares the categories of the respective fields defined from the common viewpoint in the consistency examination field pairs. The viewpoint common field search unit 120 stores the category comparison result in the comparison result storage unit 140. Thereafter, the common-view field searching unit 120 notifies the hypothesis setting unit 150 of completion of category comparison.

  The correction candidate storage unit 130 is a function for storing correction candidate information indicating that one of the consistency examination field pairs is a correction candidate. For example, a part of the storage area of the RAM 102 is used as the correction candidate storage unit 130.

  The comparison result storage unit 140 is a storage function that stores the category comparison results of the consistency examination field pairs. For example, a part of the storage area of the RAM 102 is used as the comparison result storage unit 140.

  When the hypothesis setting unit 150 receives a category comparison completion notification from the common viewpoint field search unit 120, the hypothesis setting unit 150 refers to the comparison result storage unit 140 and defines a hypothesis of a category to be added. Specifically, the hypothesis setting unit 150 detects a category that exists only in one and does not exist in the other while classifying from a common viewpoint in two different fields. Then, the hypothesis setting unit 150 makes a hypothesis that the corresponding category is added to the field where the category is insufficient. The hypothesis setting unit 150 stores the contents of the hypothesis in the hypothesis storage unit 160. Thereafter, the hypothesis setting unit 150 notifies the hypothesis verification unit 170 of completion of the hypothesis setting.

  Upon receiving the hypothesis setting completion notification from the hypothesis setting unit 150, the hypothesis verification unit 170 verifies the correctness of the hypothesis. Specifically, the hypothesis verification unit 170 uses the documents in the patent document DB to determine that the hypothesis is correct when the amount of documents to be classified into categories created by the hypothesis is a predetermined amount or more. The hypothesis verification unit 170 stores the revision plan in the revision plan storage unit 180.

The revision plan storage unit 180 is a storage function that stores the revision plan. For example, a part of the storage area of the RAM 102 or the HDD 103 is used as the revision plan storage unit 180.
The result display unit 190 displays the revision plan stored in the revision plan storage unit 180 on the monitor 11.

In the configuration as described above, F-term definition storage unit 110 stores F-term definition information in advance.
FIG. 7 is a diagram illustrating a data structure example of the F-term definition storage unit. The F term definition storage unit 110 stores an F term definition table 111. The F-term definition table 111 is provided with columns of theme code, F-term, viewpoint, first layer, second layer, and third layer.

  A field identification code (theme code) is set in the theme code column. Information in each column of the F term, viewpoint, first layer, second layer, and third layer set to the right of the theme code is the F term definition information of the corresponding theme code.

  In the F-term column, an identifier (F-term) for uniquely identifying the corresponding category in the classification system is set. F-term is the theme code with 5 characters from the left. Two characters following the theme code are an identification code of a viewpoint within the corresponding field. Two characters following the viewpoint identification code are the category identification code within the corresponding viewpoint.

  The viewpoint at the time of classification of the category indicated by the F term is set in the viewpoint column. In the column of the first hierarchy, a character string indicating the attribute of the first hierarchy to which the category indicated by the F term belongs is set. A character string indicating an attribute of the second hierarchy to which the category indicated by the F term belongs is set in the second hierarchy column. A character string indicating the attribute of the third hierarchy to which the category indicated by the F term belongs is set in the third hierarchy column. When the category indicated by the F term is the category of the first hierarchy, the columns of the second hierarchy and the third hierarchy are blank. Similarly, when the category indicated by the F term is the second hierarchy, the third hierarchy column is blank.

  In a state where such F-term definition information is stored in the F-term definition storage unit 110, the classification system revision support device 100 creates a classification system revision plan in response to a classification system revision draft creation instruction from the user. .

FIG. 8 is a flowchart showing a procedure of classification system revision proposal creation processing. In the following, the process illustrated in FIG. 8 will be described in order of step number.
[Step S11] First, the common viewpoint search section 120 searches for a field including the same viewpoint using the definition of the F term. The viewpoint common field search unit 120 sets two fields including the detected same viewpoint as a pair of fields that should be consistent (consistency examination field pair).

[Step S12] Next, the hypothesis setting unit 150 makes a hypothesis as to what revision should be made to the consistency study field pair obtained in step S11.
[Step S13] Finally, the hypothesis verification unit 170 narrows down the revision points by performing verification using the information of the gazette that is actually attached with the F term. For example, let us consider a case where there are two fields A and B in which almost the same viewpoints and categories are set. In this case, if there is a category set only in the field A, the hypothesis verification unit 170 assumes that the category can also be set under the field B. The hypothesis verification unit 170 verifies this assumption to determine whether or not to make a revised proposal.

  Here, a case where a classification system revision proposal is created based on the F-term definition table 111 shown in FIG. First, the viewpoint common field search unit 120 finds pairs of fields in which the same viewpoint and the same category are set for a huge number of F-term definition information.

  FIG. 9 is a diagram showing an example of consistency study field pairs discovered by field discovery processing including the same viewpoint. In FIG. 9, in order to compare the identity of viewpoints and the identity of categories, categories for each viewpoint in each field 51 to 54 are shown in a tree structure. For example, in the field 51 of the theme code “2B062”, a category is generated from the viewpoint of “purpose / function”. A category is also generated from the viewpoint of “purpose / function” in the field 54 of the theme code “2B063”.

  The viewpoint of “purpose / function” in the field 51 includes categories such as “improvement in operability” and “improvement in safety”. Similarly, the “purpose / function” viewpoint of the field 54 includes the categories of “improvement in operability” and “improvement in safety”. However, the category “improving planting accuracy” exists only in the field 54 and does not exist in the field 51. That is, the field 51 and the field 54 are classified from the same viewpoint, but the categories for classification from the corresponding viewpoint are only partially matched. Therefore, the common viewpoint field search unit 120 sets the consistency examination field pair for verifying the consistency of the categories of the field 51 and the field 54.

When the consistency examination field pair is determined, a hypothesis of category revision by the hypothesis setting unit 150 is established.
FIG. 10 is a diagram illustrating a hypothesis setting example. The hypothesis setting unit 150 compares whether or not the definitions of all categories match for a pair of fields 51 and 54 in which the same viewpoint and the same category are set. Here, the hypothesis setting unit 150 sets a hypothesis that, for some categories that do not match, a category defined only on one side can be defined on the other in order to match.

  In the example of FIG. 10, the definitions of categories in both fields 51 and 54 are the same for “improvement in operability” and “improvement in safety”. However, the category of “improving planting accuracy” (F-term “2B063AA08”) is set only in the field 51 of the grounding unit of the rice transplanter (theme code “2B063”). Therefore, a hypothesis is set that “improving planting accuracy” (F-term “2B063AA08”) needs to be defined in the field 54 of the rice transplanter element (theme code “2B062”).

When the hypothesis is set, the hypothesis verification unit 170 verifies the suitability of the hypothesis.
FIG. 11 is a diagram illustrating a hypothesis verification example. The hypothesis verification unit 170 verifies the suitability of the hypothesis set by the hypothesis setting unit 150. In the example of FIG. 11, the hypothesis verification unit 170 verifies the hypothesis that the category “improving planting accuracy” also needs to be defined in the field of the rice transplanter element (theme code “2B062”).

  First, the hypothesis verification unit 170 accesses the patent document DB 21 and is classified from the viewpoint of “purpose / function” (identification code “AA” of viewpoint) in the field of the grounding unit of the rice transplanter (theme code “2B063”). A set of patent documents (hypothesis source overall set 61) is acquired. Next, the hypothesis verification unit 170 extracts keywords from a set of patent documents (hypothesis source corresponding category set 62) to which a code of “improving planting accuracy” (F-term “2B063AA08”) is assigned. For example, it is assumed that keywords such as “leveling”, “sensor float”, “rice planting machine”, “sense float”, and the like are extracted. Furthermore, the hypothesis verification unit 170 extracts keywords from a set of patent documents to which “improving planting accuracy” (F-term “2B063AA08”) is not assigned. For example, it is assumed that keywords such as “leveling”, “float”, “rice planting machine”, “machine body”,... Are extracted.

  When extracting a keyword from a patent document, for example, the hypothesis verification unit 170 performs a morphological analysis of the patent document and extracts a character string of a predetermined part of speech (such as a noun or a verb) as a keyword. When a keyword list for each patent document is prepared in advance in the patent document DB 21, the hypothesis verification unit 170 can also use the keyword list.

  Further, the hypothesis verification unit 170 extracts keywords that are not included in the keywords extracted from the patent documents that do not belong to the hypothesis source corresponding category set 62 in the hypothesis source total set 61 from the keywords extracted from the hypothesis source corresponding category set 62. . The hypothesis verification unit 170 sets the extracted keyword as a feature keyword of the hypothesis source corresponding category set 62. For example, in the example of FIG. 11, keywords such as “sensor float”, “sense float”,...

  Thereafter, the hypothesis verification unit 170 accesses the patent document DB 21 and is classified in the field of rice transplanter elements (theme code “2B062”) from the viewpoint of “purpose / function” (identification code “AA” of viewpoint). A search for a set of patent documents (hypothesis target overall set 63) is performed. At this time, the hypothesis verification unit 170 searches the patent documents classified in the “other” category in the hypothesis target entire set 63. The search keyword at this time is a feature keyword of the hypothesis source corresponding category set 62. The hypothesis verification unit 170 compares the number of patent documents hit by the search (number of hits M) with a preset threshold T. If the number of hits M is greater than the threshold T, it is determined that the hypothesis is appropriate. That is, in the field of rice transplanter elements (theme code “2B062”), “improving planting accuracy” is added as a classification category from the viewpoint of “purpose / function” (identification code “AA” of viewpoint). Is recognized as appropriate.

  In this way, by verifying the set hypothesis, when there are many patent documents belonging to the newly added category, such as “improving planting accuracy”, the addition of the category is appropriate. Certified.

  As shown in FIG. 2, the category of “arithmetic processing” from the viewpoint of “purpose / function” of the rice transplanter element (theme code “2B062”) is “ Not included in the category of “purpose / function”. However, there are few inventions related to “arithmetic processing” in the field of the ground contact section of rice transplanters. Therefore, even if the category is added to the grounding part of the rice transplanter, there are few patent documents belonging to the added category. As a result, the addition of the category is judged inappropriate by the verification. As described above, even if a category peculiar to a certain field becomes a hypothesis-based category, addition of the same category to another field is excluded by verification.

  Although the example shown in FIGS. 9 to 11 shows only one hypothesis, actually, a large number of hypotheses are set from the classification system of the entire F term, and each hypothesis is verified. As a result, comprehensive improvement of a large-scale classification system can be easily performed. Hereinafter, functions and processing for reviewing the entire classification system will be described in more detail.

First, the function of the viewpoint common field search unit 120 will be described in detail.
FIG. 12 is a block diagram illustrating an internal function of the common viewpoint field search unit. The viewpoint common field search unit 120 includes a viewpoint extraction unit 121, a viewpoint identity determination unit 122, a category extraction unit 123, and a category comparison unit 124.

The viewpoint cutout unit 121 refers to the F-term definition storage unit 110 and extracts a character string indicating the viewpoint of each field.
The viewpoint identity determination unit 122 compares the character strings indicating the viewpoints of the respective fields extracted by the viewpoint cutting unit 121 with combinations of all the fields. Next, the viewpoint identity determination unit 122 determines a pair of two fields classified by the same viewpoint (consistency examination field pair) between different fields. The viewpoint identity determination unit 122 stores the consistency study field pair in the correction candidate storage unit 130. In addition, the viewpoint identity determination unit 122 counts the number of matching categories within the same viewpoint of the consistency study field pair. Then, the viewpoint identity determination unit 122 stores the number of matching categories for each consistency examination field pair in the correction candidate storage unit 130.

The category cutout unit 123 cuts out definition sentences of each category classified from the same viewpoint with respect to each field of the consistency examination field pair.
The category comparison unit 124 compares category definition sentences in each field of the consistency study field pair. At that time, the category comparison unit 124 stores the comparison result in the comparison result storage unit 140. When category comparison is completed, category comparison unit 124 notifies hypothesis setting unit 150 to that effect.

Next, the viewpoint common field search processing in the viewpoint common field search unit 120 having such an internal function will be described in detail.
FIG. 13 is the first half of the flowchart showing the procedure of the viewpoint common field search processing. In the following, the process illustrated in FIG. 13 will be described in order of step number.

[Step S <b> 21] The viewpoint cutout unit 121 refers to the F-term definition storage unit 110 and acquires the F-term definition table 111.
[Step S22] The viewpoint cutout unit 121 initializes the variable “i” to 1 and also initializes the variable “j” to 1.

  [Step S <b> 23] The viewpoint cutout unit 121 acquires the i-th theme code of the F-term definition table 111. The viewpoint cutout unit 121 sets the acquired theme code to the variable “Ti”. The variable “Ti” indicates a field (reference field) that serves as a reference source when setting a hypothesis.

  [Step S24] The viewpoint cutout unit 121 acquires the j-th theme code in the F-term definition table 111. The viewpoint cutout unit 121 sets the acquired theme code to the variable “Tj”. The variable “Tj” indicates a field to be added to a category (revision target field) when setting a hypothesis.

  [Step S25] The viewpoint cutout unit 121 compares the values of the current variable i and the variable j, and determines whether or not the values are different (i ≠ j). If the values are different, the process proceeds to step S27. If the values are equal, the process proceeds to step S26.

  [Step S26] When the values of the variable i and the variable j are equal, the viewpoint cutout unit 121 increments the variable j and advances the process to step S24. Thus, when “j = i”, the same theme code is compared, so that the comparison process is skipped and j is incremented.

  [Step S27] If the values of the variable i and j are not equal, the viewpoint extraction unit 121 corresponds to the fields indicated by the variable “Ti” and the variable “Tj” on the F-term definition table 111, respectively. F-term definition information is acquired.

[Step S28] The viewpoint cutout unit 121 extracts the viewpoint of the field indicated by the variable “Ti” and the viewpoint of the field indicated by the variable “Tj” from the acquired F-term definition information.
[Step S29] The viewpoint identity determination unit 122 detects a common viewpoint among the viewpoint of the field indicated by the variable “Ti” and the viewpoint of the field indicated by the variable “Tj”. Specifically, when the character strings indicating the viewpoints match, it is determined that the viewpoints are common.

  [Step S30] The viewpoint identity determination unit 122 counts the number of matching categories of common viewpoints. Specifically, the viewpoint identity determination unit 122 compares the categories in each field related to the common viewpoints, and counts the number of categories having the same definition content.

  [Step S <b> 31] The viewpoint identity determination unit 122 adds matching viewpoint information to the correction candidate table in the correction candidate storage unit 130. Specifically, the viewpoint identity determination unit 122 adds the theme codes of the two fields to be compared, the matching viewpoints, and the number of matching categories to the correction candidate table. If there is no commonality between the viewpoint of the theme code indicated by the variable “Ti” and the viewpoint of the theme code indicated by the variable “Tj”, no information is added to the correction candidate table. Thereafter, the process proceeds to step S41 (see FIG. 15).

  FIG. 14 is a diagram illustrating an example of a correction candidate table in the correction candidate storage unit. As illustrated in FIG. 14, the correction candidate storage unit 130 includes a correction candidate table 131. The correction candidate table 131 includes columns for theme code A, theme code B, matching viewpoints, number of matching categories, and priority. Information arranged in the horizontal direction of each column is information indicating one correction candidate.

  In the theme code A column, the theme code of the reference field that is the reference source of the hypothesis is set. Specifically, the theme code set in the variable “Ti” is set in the field of the theme code A. In the field of the theme code B, the theme code of the amendment target field to which the theme code is added according to the hypothesis is set. Specifically, the theme code set in the variable “Tj” is set in the field of the theme code B. A pair of the theme code field set in the theme code A and the theme code field set in the theme code B is a consistency examination field pair.

  In the matching viewpoint column, a viewpoint that matches the field set in the theme code A and the field set in the theme code B is set. In the column of the number of matching categories, the number of matching categories for classification from the viewpoint of matching is set. In the priority column, a priority for maintaining the consistency of the corresponding consistency examination field pair is set. Note that the priority column is blank at the stage of processing shown in FIG.

FIG. 15 is the second half of the flowchart showing the procedure of the viewpoint common field search processing. In the following, the process illustrated in FIG. 15 will be described in order of step number.
[Step S41] The category extraction unit 123 extracts category definition sentences defined in each field for the viewpoints that match the field indicated by the variable “Ti” and the field indicated by the variable “Tj”. To do.

  [Step S42] The category comparison unit 124 extracts keywords from the definition sentence extracted in step S41. At this time, the category comparison unit 124 uses the extracted keyword to determine the identity of the category, and therefore deletes character strings that do not characterize the content of the category, such as particles and inflections.

  [Step S43] The category comparison unit 124 compares keywords between the two selected categories for all combinations obtained by selecting one category from each of the two fields to be compared, and matches or does not match the keywords. Determine.

  [Step S44] The category comparison unit 124 creates a comparison table corresponding to the consistency study field pair currently being processed in the comparison result storage unit 140, and writes the comparison result in the comparison table. Specifically, if all of the keywords match, the category comparison unit 124 writes a comparison result indicating matching. Moreover, the category comparison part 124 will write the comparison result which shows a partial match, if only a part of keywords match. Further, the category comparison unit 124 writes a comparison result indicating mismatch when there is no matching keyword.

[Step S <b> 45] The category comparison unit 124 calculates the priority from the number of categories having the same viewpoint in common between the two fields to be compared. At this time, the priority is set higher because the consistency should be maintained as the proportion of the matching categories in the common viewpoints of the two fields indicated by the variable “Ti” and the variable “Tj” increases. For example, the category comparison unit 124 calculates the priority with the following formula.
Priority = S / ((Ai + Aj) / 2) (1)
Here, S is the number of coincidence of categories respectively defined in the two fields with respect to the viewpoint common to the two fields indicated by the variable “Ti” and the variable “Tj”. Ai is the total number of categories defined in the corresponding viewpoint in the field indicated by the variable “Ti” regarding the viewpoint common to the two fields. Aj is the total number of categories defined in the corresponding viewpoint in the field indicated by the variable “Tj” with respect to the viewpoint common to the two fields.

  For example, assume that a category as shown in FIG. 2 is defined. In this case, there are five matching categories: “Improved performance of transplanter etc.” “Improved operability” “Improved safety” “Improved work accuracy” “Reduced size / weight or rationalized manufacturing”. The total number of categories set for the purpose / function of the rice transplanter element (2B062) is 12 categories. Further, 13 categories are set from the viewpoint of the purpose / function of the grounding unit (2B063) of the rice transplanter. The priority is then 0.4.

The category comparison unit 124 that has calculated the priority writes the number of matching categories and the priorities of each consistency examination field pair in the correction candidate table 131.
Note that when the categories defined in the common viewpoint are completely the same in the two comparison target fields, the priority is “1”. In this case, since it can be said that the consistency is already maintained, the category comparison unit 124 deletes the corresponding consistency examination field pair from the correction candidate table 131.

  [Step S46] The category comparison unit 124 determines whether or not the variable “j” is equal to the number of theme codes registered in the F-term definition table. If the variable “j” has reached the number of theme codes, the process proceeds to step S48. If the variable “j” is less than the number of theme codes, the process proceeds to step S47.

[Step S47] If the variable “j” is less than the number of theme codes, the category comparison unit 124 increments j and advances the process to step S24.
[Step S48] If the variable “j” has reached the number of theme codes, the category comparison unit 124 determines whether the variable “i” is equal to the number of theme codes registered in the F-term definition table. . When the variable “i” has reached the number of theme codes, the viewpoint common field search process ends. If the variable “i” is less than the number of theme codes, the process proceeds to step S49.

[Step S49] If the variable “i” is less than the number of theme codes, the category comparison unit 124 increments i and advances the process to step S23.
As described above, it is possible to extract a consistency study field pair for which a common viewpoint is set. The comparison result storage unit 140 stores a comparison table for each consistency examination field pair.

  FIG. 16 is a diagram illustrating an example of a comparison table in the comparison result storage unit. The comparison result storage unit 140 stores comparison tables 141, 142, 143,... For each consistency study field pair. For example, the comparison table 141 corresponds to the consistency examination field pair between the field (theme code “2B063”) of the “rice transplanter grounding portion” and the field of the “rice transplanter element” (theme code “2B062”).

  The creation process of the comparison table 141 is divided into a tabular data structure creation process and a comparison result writing process for the data structure. FIG. 16 shows a comparison table 141 after the creation of a tabular data structure.

  The category comparison unit 124 creates a tabular comparison table 141. Specifically, the category comparison unit 124 sets the F term of each category of the theme code A regarding the matching viewpoint of the consistency study field pairs as the column label of the comparison table 141. Further, the category comparison unit 124 sets the definition of the category corresponding to the F term set in the column label below the column label.

  Further, the category comparison unit 124 sets the F term of each category of the theme code B related to the matching viewpoint of the consistency examination field pairs as the row label of the comparison table 141. Further, the category comparison unit 124 sets a category definition corresponding to the F-term set in the row label to the right of the row label.

Under the last line in which the F term of the theme code B is set, the characters “total” are set as the line label.
The result of comparing the categories set in the matching viewpoint of the consistency study field pairs is written in the cells of the comparison table 141 created in this way.

  FIG. 17 is a diagram illustrating the comparison table after the determination result is written. As a comparison result, “◯” is written when the definition keywords are completely matched, “Δ” is written when they are partially matched, and “X” is written when they are not matched. In the example of FIG. 17, “operability improvement” and “stability improvement” categories are completely the same. In addition, “improvement in running performance” and “balance and stability” in the reference field (theme code A) partially coincide with “improvement in running ability or stability” in the revision target field (theme code B). Note that when the comparison result is registered, the “total” column is left blank.

Thereafter, the category comparison unit 124 calculates the priority for each consistency examination field pair and sets the priority in the correction candidate table 131.
FIG. 18 is a diagram illustrating a correction candidate table after priority setting. As shown in FIG. 18, a priority is set for each consistency study field pair. The higher the priority value, the higher the need to maintain consistency.

  When the comparison table 141, 142, 143,... For each consistency examination field pair as shown in FIG. 17 and the correction candidate table 131 as shown in FIG. Setting processing is executed.

FIG. 19 is a block diagram illustrating the internal functions of the hypothesis setting unit. The hypothesis setting unit 150 includes a comparison table totaling unit 151 and a hypothesis target extraction unit 152.
The comparison table totaling unit 151 totals information related to the match / mismatch of the categories of the comparison tables 141, 142, 143,... In the comparison result storage unit 140. At that time, the comparison table totaling unit 151 refers to the priority set in the correction candidate table 131 in the correction candidate storage unit 130, and totals the corresponding comparison tables in order from the consistency examination field pair having the highest priority. I do.

The hypothesis target extraction unit 152 determines a category to be added to the revision target field based on the tabulation result, and creates a hypothesis table. The hypothesis table is stored in the hypothesis storage unit 160.
Next, the hypothesis setting process will be described in detail.

FIG. 20 is a flowchart illustrating the procedure of hypothesis setting processing. In the following, the process illustrated in FIG. 20 will be described in order of step number.
[Step S51] The comparison table totaling unit 151 sorts the records in the correction candidate table 131 in descending order of priority. A consistency study field pair with a high priority is originally highly consistent. If high consistency has already been obtained, the categories of each field can be matched with a small amount of correction. Therefore, a candidate with a high priority is processed first.

[Step S <b> 52] The comparison table totaling unit 151 selects the consistency study field pair having the highest priority in the correction candidate table 131.
[Step S53] The comparison table totaling unit 151 totals the comparison tables corresponding to the selected consistency examination field pairs in the vertical direction. If there is a “○” in the vertical direction, a category that matches the category of the reference field (the field indicated by “Theme Code A”) set in the label of the corresponding column has already been included in the amendment target field. It will exist. In this case, it is not necessary to add a new category to the revision target field (the field indicated by “theme code B”). Therefore, if there is at least one “◯” in the vertical direction, the comparison table totaling unit 151 writes a symbol “◯” indicating that it matches in the total column of the corresponding column.

  Further, when all of the comparison tables in the vertical direction are “x”, it means that there is no category corresponding to the category in the reference field in the revision target field. In this case, the hypothesis is that a new category is created under the revision target field as a revision proposal for the classification system. Therefore, the comparison table totaling unit 151 writes the symbol “X” indicating that the category is the addition target category in the totaling column if all the “×” in the vertical direction.

  Furthermore, when “Δ” is present when tallying in the vertical direction, there is a partially matching category. In the present embodiment, partially matching categories are not subject to correction. However, in the case of partial matching, the categories can be more accurately matched by amending the division or integration of categories. Therefore, the comparison table totalization unit 151 indicates that the category is a division or integration target category in the total column of the corresponding column if there is no “◯” in the vertical direction and at least one “Δ”. Write the symbol “△”.

  [Step S54] When the comparison table aggregation unit 151 completes the aggregation of the comparison table that is the aggregation object, the hypothesis target extraction unit 152 selects one unselected aggregation result in the aggregation column.

  [Step S55] The hypothesis target extraction unit 152 determines whether or not the counting result is “x”. If the count result is “x”, the process proceeds to step S56. If the count result is other than “x”, the process proceeds to step S57.

  [Step S56] If the selected aggregation result is “x”, the hypothesis target extraction unit 152 selects a hypothesis that the category of the reference field set in the column of the aggregation result is added to the revision target field. And registered in the hypothesis table in the hypothesis storage unit 160. In this way, a new hypothesis is written in the hypothesis table only when a new category needs to be created in the theme code.

  [Step S57] The hypothesis target extraction unit 152 determines whether all the tabulation results have been selected. If all the tabulation results have been selected, the process proceeds to step S58. If there is an unselected count result, the process proceeds to step S54, and another count result is selected.

  [Step S58] The comparison table totaling unit 151 determines whether or not all consistency examination field pairs in the correction candidate table have been selected. When the selection of all consistency study field pairs is completed, the hypothesis setting process ends. If there is an unselected consistency examination field pair, the process proceeds to step S52, and the next consistency examination field pair is selected.

In this way, the comparison table is aggregated, and a hypothesis is set by the hypothesis setting unit 150 based on the aggregation result.
FIG. 21 is a diagram showing an example of the tabulated comparison table. In the comparison table 141, the category match / mismatch determination results are tabulated for each column. In the present embodiment, a hypothesis is established as adding a category of “×” as the total result to the revision target field. The established hypothesis is set in the hypothesis table.

  FIG. 22 is a diagram illustrating a data structure example of a hypothesis table in the hypothesis storage unit. The hypothesis table 161 in the hypothesis storage unit 160 includes columns for a reference field, a field to be revised, a viewpoint, a hypothesis, and a verification result.

  The theme code of the hypothetical reference field is set in the reference field column. The theme code of the revision target field of the hypothesis is set in the field of the revision target field. In the viewpoint column, a viewpoint to which consistency is to be maintained is set. In the hypothesis column, the change contents (hypotheses) of the classification system to be applied to the revision target field are set. For example, adding a predetermined category is set as a hypothesis. In the verification result column, a verification result indicating whether or not the hypothesis is appropriate is set. Note that the verification result column is blank after the hypothesis setting and before verification.

When the hypothesis setting process is completed, the hypothesis verification unit 170 performs hypothesis verification.
FIG. 23 is a block diagram illustrating an internal function of the hypothesis verification unit. The hypothesis verification unit 170 includes a feature keyword extraction unit 171, a feature keyword storage unit 172, a patent document search unit 173, and a hypothesis determination unit 174.

  The feature keyword extraction unit 171 acquires from the patent document DB 21 patent documents classified in the hypothesis reference field set in the hypothesis table 161 in the hypothesis storage unit 160. Then, the feature keyword extraction unit 171 extracts feature keywords of patent documents belonging to the category to be added based on the hypothesis. The feature keyword extraction unit 171 stores the extracted feature keyword in the feature keyword storage unit 172.

The feature keyword storage unit 172 is a storage function that stores feature keywords. For example, a part of the storage area of the RAM 102 is used as the feature keyword storage unit 172.
The patent document search unit 173 searches for patent documents classified in the hypothesis revision target field in the patent document DB 21 based on the feature keywords of the patent documents belonging to the category to be added by the hypothesis.

  The hypothesis determination unit 174 determines whether the hypothesis is appropriate according to the search result by the patent document search unit 173. When the hypothesis determination unit 174 determines that the hypothesis is appropriate, the hypothesis determination unit 174 stores the content of the hypothesis in the revision plan storage unit 180.

Next, the procedure of the hypothesis verification process will be described in detail.
FIG. 24 is a flowchart showing the procedure of hypothesis verification processing. In the following, the process illustrated in FIG. 24 will be described in order of step number.

[Step S61] The feature keyword extraction unit 171 selects one unprocessed hypothesis in the hypothesis table 161.
[Step S62] The feature keyword extraction unit 171 performs a feature keyword extraction process for an additional category. Details of this processing will be described later (see FIG. 25).

  [Step S63] The patent document search unit 173 searches the patent keywords classified as “others” with the feature keyword of the category to be added from the viewpoint of achieving consistency in the revision target field of the patent document DB 21. At this time, when there are a plurality of feature keywords, a logical sum (OR) search of each feature keyword is performed.

[Step S64] The patent document search unit 173 substitutes the variable Z for the number of hits by the search.
[Step S65] The hypothesis determination unit 174 determines whether or not the number of hits set in the variable Z is equal to or greater than a threshold value. If the number of hits is equal to or greater than a preset threshold, the process proceeds to step S66. If the number of hits is less than the threshold, the process proceeds to step S67.

  The threshold may be changed according to the number of patent documents registered in the patent document DB 21. For example, in the hypothesis determination unit 174, a threshold value of the hit rate with respect to the total number of patent documents registered as search targets is defined in advance. The hypothesis determination unit 174 determines that the number of hits is equal to or greater than the threshold if the number of hits is equal to or greater than a value obtained by multiplying the total number of patent documents by a threshold.

  [Step S66] If the number of hits is equal to or greater than the threshold, the hypothesis determination unit 174 sets a symbol “◯” indicating that the hypothesis is appropriate in the verification result column of the selected hypothesis. Thereafter, the process proceeds to step S68.

[Step S67] If the number of hits is less than the threshold, the hypothesis determination unit 174 sets a symbol “x” indicating that the hypothesis is inappropriate in the verification result column of the selected hypothesis.
[Step S68] The feature keyword extraction unit 171 determines whether or not all hypotheses have been verified. If all the hypotheses have been verified, the process proceeds to step S69. If there is an unprocessed hypothesis, the process proceeds to step S61 and the next hypothesis is selected.

  [Step S69] The hypothesis determination unit 174 extracts the revision target field, viewpoint, and hypothesis from the hypothesis in which the verification result column of the hypothesis table 161 is “◯”, and generates a revision plan table. The generated revision plan table is stored in the revision plan storage unit 180.

Next, the feature keyword extraction process of the additional category will be described in detail.
FIG. 25 is a flowchart showing a procedure of additional keyword feature keyword extraction processing. In the following, the process illustrated in FIG. 25 will be described in order of step number.

  [Step S71] The feature keyword extraction unit 171 accesses the patent document DB 21. The feature keyword extraction unit 171 then selects a keyword set α from a set of patent documents classified into a category other than the category that is assumed to be added to the revision target field among the categories in which the reference field is classified from the same viewpoint as the revision target field. To extract. For extracting the keyword set α, a morphological analysis technique, a compound word extraction technique, and an information extraction technique, which are generally known as existing techniques, can be used.

  [Step S72] The characteristic keyword extraction unit 171 accesses the patent document DB 21. Then, the feature keyword extraction unit 171 extracts the keyword set β from the set of patent documents classified in the category that is assumed to be added to the revision target field among the categories in which the reference field is classified from the same viewpoint as the revision target field. .

  [Step S73] The feature keyword extraction unit 171 statistically processes the keyword set α and the keyword set β, and extracts a feature keyword set δ characteristic of the keyword set β. As a statistic, a chi-square value, a cullback / librar information amount, or the like can be used.

  In the feature keyword set δ, it is also possible to exclude a keyword whose weight (degree of feature display) is large only in one of the entire reference field and the entire revision target field. As a result, it is possible to prevent an erroneous result from being produced by a keyword specific to each field in the reference field and the field to be revised.

Here, the process of extracting the feature keyword set δ using the chi-square value and the amount of information of the cullback / librar is described in detail.
FIG. 26 is a diagram showing the classification status of patent documents in the reference field. In this example, it is assumed that “field A” is the reference field. Further, a category assumed to be added to the revision target field is “category c”. Patent Document Number belonging to the entire "field A" is "N _A".
The number of patent documents belonging to “category c” is “N _C ”.

Here, the weight of the keyword K is calculated. In this case, a search using the keyword K is performed for the entire “field A”. As a result, the number of patent documents K _A including the keyword K in the entire “field A” is obtained. In addition, a search using a keyword K is performed for patent documents belonging to “category c”. Thus, the number of patent documents K _C including the keyword K among the patent documents belonging to “category c” is obtained.

The calculation formula of the chi-square value (χ ² ) in this case is as follows.
χ ² = (K _C −E) ² / E (2)
Here, E is an expected value of the keyword K in the entire field A. At this time, the expected value E of the keyword K can be calculated as follows.
E = (K _A × N _C ) / N _A (3)
The larger the value of χ ² obtained by the equations (2) and (3), the stronger the keyword K represents the feature of “category c”.

Further, under the classification situation shown in FIG. 26, Kullback-Leibler divergence D _KL can be calculated by the following equation.
D _KL = P × log (P / Q) (4)
Here, P is the probability of including the keyword K among the patent documents included in the “category c”. Q is a probability of including the keyword K in the entire “field A”. P and Q are represented by the following equations, respectively.
P = K _C / N _C (5)
Q = K _A / N _A (6)
The larger the value of the Cullback / librar information amount D _KL obtained by the equations (4) to (6), the stronger the keyword K represents the feature of the “category c”.

  In this way, using the chi-square value and the Cullback / librar information amount, it is possible to calculate the weight related to an arbitrary keyword (how strongly the feature of “category c” is expressed). The feature keyword extraction unit 171 obtains a weight for each keyword included in the keyword set extracted from the patent document belonging to “category c”. Then, the feature keyword extraction unit 171 sets, for example, a set of keywords whose weight is greater than a predetermined value as a feature keyword set δ.

Each keyword included in the feature keyword set δ is stored in a feature keyword table provided in the RAM 102.
FIG. 27 is a diagram illustrating an example of the data structure of the feature keyword table. The feature keyword storage unit 172 stores a feature keyword table 172a. The feature keyword table 172a has columns for F-term, category, and feature keyword.

  The F term of each category is set in the F term column. The category definition content is set in the category column. In the feature keyword column, the feature keyword of the corresponding category is set.

The hypothesis is verified based on the feature keyword of each category, and the suitability of the hypothesis is set in the verification result column of the hypothesis table 161.
FIG. 28 is a diagram illustrating an example of a hypothesis table after completion of hypothesis verification. In the example of FIG. 28, the symbol “○” indicating that it is appropriate for the hypothesis of adding “improving planting accuracy” to the category of classification in the field of “purpose / function” in the field of the theme code “2B062”. Is set. On the other hand, the symbol “x” indicating that the lamp is inappropriate is set for the hypothesis of adding “lamp” to the category of classification from the viewpoint of “light source” in the field of the theme code “2F065”.

Only appropriate hypotheses are extracted from the hypothesis table 161 in which such verification results are registered, and stored in the revision plan storage unit 180 as revision plans.
FIG. 29 is a diagram illustrating an example of the data structure of the revision plan storage unit. The revision plan storage unit 180 stores a revision plan table 181 for registering revision plans. The revision plan table 181 includes columns for a theme code, a viewpoint, and a revision plan.

  The theme code of the field to be revised is set in the theme code column. In the viewpoint column, a viewpoint to which a category is added is set. The definition of the category to be added is set in the revision column.

  As described above, it is possible to create an appropriate revision proposal for the classification category based on the viewpoint. The content of the revision plan stored in the revision plan table 181 is displayed on the monitor 11 by the result display unit 190. The user can refer to the revision plan displayed on the monitor 11 and can use it for the revision of the F-term classification system.

  In addition, the hypothesis is verified using the patent documents registered in the patent document DB 21. Since patent document DB21 has already classified a huge amount of patent documents by F-term, the suitability of the hypothesis can be accurately determined by using patent document DB21.

[Second Embodiment]
Next, a second embodiment will be described. In the second embodiment, verification is performed using a classifier based on machine learning. The processing function in the second embodiment is different from the first embodiment only in the function of the hypothesis verification unit.

  FIG. 30 is a block diagram illustrating an internal function of the hypothesis verification unit according to the second embodiment. The hypothesis verification unit 170a in the second embodiment includes a common keyword extraction unit 175, a common keyword storage unit 176, a machine learning unit 177, a classification unit 178, and a hypothesis determination unit 179.

  The common keyword extraction unit 175 accesses the patent document DB 21 and extracts keywords appearing in both the hypothesis reference field and the revision target field stored in the hypothesis storage unit 160. The common keyword extraction unit 175 stores the set of extracted common keywords (common keyword set) in the common keyword storage unit 176.

  The common keyword storage unit 176 is a storage function that stores a common keyword set for each hypothesis. For example, a part of the storage area of the RAM 102 is used as the common keyword storage unit 176.

The machine learning unit 177 performs machine learning based on the common keyword set using patent documents classified in the reference field, and generates a classifier related to the category to be added.
The classifying unit 178 classifies patent documents belonging to the “others” category from the viewpoint of consideration of the revision target field, using a classifier generated by machine learning.

The hypothesis determination unit 179 is a classifier newly created by machine learning, and determines whether the hypothesis is appropriate based on the number of patent documents classified into categories to be added.
FIG. 31 is a flowchart illustrating a procedure of hypothesis verification processing using machine learning. In the following, the process illustrated in FIG. 31 will be described in order of step number.

[Step S81] The common keyword extraction unit 175 selects one unprocessed hypothesis in the hypothesis table 161.
[Step S82] The common keyword extraction unit 175 extracts keywords that appear in both the reference field and the revision target field of the selected hypothesis, and sets them as a common keyword set γ. The generated common keyword set γ is stored in the common keyword storage unit 176.

  [Step S83] The machine learning unit 177 weights each keyword in the common keyword set γ by machine learning. Specifically, the machine learning unit 177 uses, as a positive example, patent documents classified in an additional candidate category for the revision target field in the reference field of the selected hypothesis. In addition, the machine learning unit 177 uses, as a negative example, patent documents that are classified in categories other than the category to be added to the revision target field in the reference field of the selected hypothesis. The machine learning unit 177 calculates the weight of each keyword so that the more the keyword is included in the positive example and the more the keyword is not included in the negative example, the greater the weight of the keyword. The weight information for each keyword is a classifier for the category to be added.

  [Step S84] The classification unit 178 uses the classifier to classify whether or not the patent document classified as “Other” in the classification from the viewpoint of examination in the revision target field of the selected hypothesis belongs to the additional candidate category. To do. Specifically, the classification unit 178 extracts keywords from the patent documents to be classified and collates them with the classifier. It is determined that there is a high possibility that the patent document belongs to the additional candidate category as the number of keywords set with a high weight in the classifier is detected from the patent documents to be classified. The classification unit 178 classifies the patent documents that are determined to be included in the additional candidate category with a possibility greater than or equal to a predetermined value into the category.

[Step S85] The classification unit 178 substitutes the variable Z for the number of patent documents classified in the additional candidate category.
[Step S86] The hypothesis determination unit 179 determines whether or not the number of cases set in the variable Z is equal to or greater than a threshold value. If the number is equal to or greater than a preset threshold, the process proceeds to step S87. If the number is less than the threshold, the process proceeds to step S88.

  The threshold may be changed according to the number of patent documents registered in the patent document DB 21. For example, the hypothesis determination unit 179 predefines a threshold value of the relevance ratio with respect to the total number of patent documents registered as classification targets. The hypothesis determination unit 174 determines that the number of cases classified into the additional candidate category is equal to or greater than the threshold if the number of cases classified into the additional candidate category is equal to or greater than the value obtained by multiplying the total number of patent documents by the threshold.

  [Step S87] If the number of cases classified into the additional candidate category is equal to or greater than the threshold, the hypothesis determination unit 179 sets a symbol “◯” indicating that the hypothesis is appropriate in the verification result column of the selected hypothesis. Thereafter, the process proceeds to step S89.

  [Step S88] If the number of cases classified into the additional candidate category is less than the threshold, the hypothesis determination unit 179 sets a symbol “x” indicating that the hypothesis is inappropriate in the verification result column of the selected hypothesis. .

  [Step S89] The common keyword extraction unit 175 determines whether or not all hypotheses have been verified. If all hypotheses have been verified, the process proceeds to step S90. If there is an unprocessed hypothesis, the process proceeds to step S81 and the next hypothesis is selected.

  [Step S90] The hypothesis determination unit 179 extracts the revision target field, viewpoint, and hypothesis from the hypothesis in which the verification result column of the hypothesis table 161 is “◯”, and generates a revision plan table. The generated revision plan table is stored in the revision plan storage unit 180.

  In this way, the number of patent documents belonging to the additional candidate category can be predicted using machine learning. If machine learning is used, it is possible to appropriately classify patent documents at a high level, and it is also possible to more appropriately determine whether or not to add an additional candidate category set by a hypothesis.

[Other application examples]
The feature keyword extraction technique can be modified as follows.
The feature keyword is limited to a keyword for distinguishing the non-addition candidate category and the addition candidate category in the reference field. Therefore, in the feature keyword extraction unit 171, a difference keyword between a keyword extracted from a patent document belonging to a non-addition candidate category and a keyword extracted from a patent document belonging to an addition candidate category is used as a feature keyword.

  Prevent erroneous results from keywords that are unique to each of the standard and revised areas. For this purpose, the feature keyword extraction unit 171 adds a keyword whose weight is increased only in one of the appearance keyword in the entire field A and the appearance keyword in the entire field B among the obtained feature keywords. Suppose that it is not a feature keyword of a candidate category.

  Emphasize keywords that reflect the perspective of classification. Therefore, in the feature keyword extraction unit 171, the obtained feature keywords hardly appear in the defined categories under the amendment target field, and among the categories representing “others” under the amendment target field. It is used as a feature keyword only when it hits more than a certain amount.

  In the above-described embodiment, a field with high consistency is preferentially paired with a consistency examination field. Therefore, it is possible to place the revision plan on the revision plan that has been preferentially studied and present it to the user in order from the higher revision plan. Thereby, the user can know the location of the classification system which should be revised preferentially.

  Further, when the hypothesis is verified, the number of patent documents that are expected to be included in a new category when the revision is set is obtained. Therefore, when the revision plan table is presented to the user, the number of cases included in the new category can be added. Thereby, the user can appropriately determine whether or not a category needs to be added.

  In addition, when a user inputs an instruction to add a category indicated in the amendment, a new category classification code is automatically added to a patent document belonging to the added category in the patent document DB 21. You can also.

Moreover, although the said embodiment demonstrated the improvement of the classification system of a patent document, it is applicable to the various classification systems which classify | categorize information according to a viewpoint.
FIG. 32 is a diagram illustrating an example of a classification system of information related to cooking recipes. In the cooking recipe classification system shown in FIG. 32, fields are defined according to cooking types such as “vegetable side dishes”, “salad”, “bread”, and “sweets”. In each field, categories are defined in terms of “material”, “cooking method”, and “calorie”. In addition to the viewpoint shown in FIG. 32, categories can also be defined from the viewpoint of “material cost”.

  Here, if the classification system revision support apparatus according to the first and second embodiments is used, for example, “confectionery” is set as the reference field, and “vegetables” as the category of “cooking” is “pan”. Can be set and verified.

  FIG. 33 is a diagram illustrating a classification system of exhibition information related to auctions via a network. In the auction exhibition information classification system shown in FIG. 33, fields are defined by product types such as “air conditioner”, “clothes”, “bag”, and “personal computer”. In each field, categories are defined in terms of “manufacturer”, “price range”, and “CPU”. In addition to the viewpoints shown in FIG. 33, categories can be defined from the viewpoints of “specifications” and “days until the end of the auction”.

  Here, if the classification system revision support device according to the first and second embodiments is used, for example, “PC” is set as a reference field, and “N Denki” is set as a category of “manufacturer” in “air conditioner”. You can set and verify additional hypotheses.

  In addition to those shown in FIGS. 32 and 33, the above-described embodiment can be applied to, for example, a restaurant information classification system and music information classification system. If it is a classification system of information about restaurants, it is possible to divide fields by genres such as Japanese, Chinese and Italian, and to define categories in terms of areas and price ranges in each field. And these categories can be improved by the function of the said embodiment. If it is a classification system of information about music, categories such as pops, classical music and jazz can be defined from the viewpoint of country, price range, singer's gender and group composition. And these categories can be improved by the function of the said embodiment.

  The above processing functions can be realized by a computer. In that case, a program describing the processing contents of the functions that the classification system revision support device should have is provided. By executing the program on a computer, the above processing functions are realized on the computer. The program describing the processing contents can be recorded on a computer-readable recording medium. Examples of the computer-readable recording medium include a magnetic storage device, an optical disk, a magneto-optical recording medium, and a semiconductor memory. Examples of the magnetic storage device include a hard disk device (HDD), a flexible disk (FD), and a magnetic tape. Optical discs include DVD, DVD-RAM, CD-ROM / RW, and the like. Magneto-optical recording media include MO (Magneto-Optical disc).

  When distributing the program, for example, a portable recording medium such as a DVD or a CD-ROM in which the program is recorded is sold. It is also possible to store the program in a storage device of a server computer and transfer the program from the server computer to another computer via a network.

  The computer that executes the program stores, for example, the program recorded on the portable recording medium or the program transferred from the server computer in its own storage device. Then, the computer reads the program from its own storage device and executes processing according to the program. The computer can also read the program directly from the portable recording medium and execute processing according to the program. Further, each time the program is transferred from the server computer, the computer can sequentially execute processing according to the received program.

  In addition, at least a part of the above processing functions can be realized by an electronic circuit such as a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), or a PDL (Programmable Logic Device).

  As mentioned above, although embodiment was illustrated, the structure of each part shown by embodiment can be substituted by the other thing which has the same function. Moreover, other arbitrary structures and processes may be added. Further, any two or more configurations (features) of the above-described embodiments may be combined.

The main technical features of the embodiment described above are as follows.
(Supplementary note 1)
Refers to a classification system information storage means for storing classification system information indicating a classification system in which a classification viewpoint is defined in each of a plurality of fields and a category for classification according to the defined viewpoint is provided. Extract at least two fields with categories,
Compare the categories established in the common viewpoint of each extracted field,
Define a hypothesis that a category that exists only in some of the extracted fields is an additional target category, and the additional target category is to be added to another field,
Referring to a database in which information to be classified by the classification system is stored, and when the additional target category is added according to the hypothesis, additional category conforming information that belongs to the additional target category is detected,
Determining the suitability of the hypothesis based on the amount of additional category conformance information detected;
Outputting the hypothesis determined to be appropriate as a revised draft,
Classification system revision support program characterized by having processing executed.

  (Supplementary Note 2) When a plurality of combinations of at least two fields in which categories are provided from a common viewpoint are extracted, the number of categories whose definitions match among the categories provided from the common viewpoints in each field The classification system revision support program according to appendix 1, wherein the higher the number of combinations, the higher the priority of hypothesis setting processing.

(Supplementary Note 3) When detecting the additional category conformance information,
Among the extracted fields, a field that already has the additional target category as a category according to the common viewpoint is set as a reference field, and a field that does not have the additional target category as a category according to the common viewpoint As the revision target field, the information belonging to the reference field in the database is analyzed to extract the additional target category feature information indicating the characteristics of the information belonging to the additional target category, and each of the revision target fields in the database It is determined whether each piece of information belonging to the revision target field corresponds to the additional category conforming information depending on whether the information includes the additional target category feature information.
The classification system revision support program according to appendix 1, wherein the program is executed.

(Supplementary Note 4) When extracting the additional target category feature information,
Information that belongs to the additional target category by comparing a keyword obtained from information belonging to the additional target category among information belonging to the reference field in the database and a keyword obtained from information not belonging to the additional target category A feature keyword indicating a feature of the feature, and the feature keyword as the addition target category feature information,
The classification system revision support program according to appendix 3, wherein the processing is executed.

(Supplementary Note 5) When extracting the additional target category feature information,
A classifier obtained by performing machine learning using information belonging to the additional target category as a positive example and information not belonging to the additional target category as negative examples among information belonging to the reference field in the database, and obtained as a result of machine learning The additional target category feature information
The classification system revision support program according to appendix 3, wherein the processing is executed.

  (Additional remark 6) When defining the said hypothesis, the character string which defines the category created from the common viewpoint of each extracted field is analyzed, the keyword which shows the characteristic of each category is extracted, Appendix 1 characterized in that if a keyword extracted from a character string defining a category is not extracted from any category in another field, the category from which the keyword is extracted is set as an addition target category. The support system revision support program described.

(Supplementary note 7) Refer to a classification system information storage unit that stores classification system information indicating a classification system in which a classification viewpoint is defined in each of a plurality of fields and a category for classification according to the defined viewpoint is provided, A viewpoint common field search means for extracting at least two fields having categories in a common viewpoint;
Compare the categories provided from the common viewpoint of each extracted field, and select categories that exist only in some of the extracted fields as additional target categories, and set the additional target categories as other fields. A hypothesis definition means for defining a hypothesis to be added to
Referring to a database in which information to be classified by the classification system is stored, when the additional target category is added according to the hypothesis, additional category conforming information belonging to the additional target category is detected and detected Hypothesis verification means for determining the suitability of the hypothesis based on the amount of the additional category conforming information that has been determined, and outputting the hypothesis determined to be appropriate as a revision plan;
A classification system revision support device characterized by comprising:

(Appendix 8) The computer
Refers to a classification system information storage means for storing classification system information indicating a classification system in which a classification viewpoint is defined in each of a plurality of fields and a category for classification according to the defined viewpoint is provided. Extract at least two fields with categories,
Compare the categories established in the common viewpoint of each extracted field,
Define a hypothesis that a category that exists only in some of the extracted fields is an additional target category, and the additional target category is to be added to another field,
Referring to a database in which information to be classified by the classification system is stored, and when the additional target category is added according to the hypothesis, additional category conforming information that belongs to the additional target category is detected,
Determining the suitability of the hypothesis based on the amount of additional category conformance information detected;
Outputting the hypothesis determined to be appropriate as a revised draft,
Classification system revision support method characterized by this.

1 Classification system information storage means 2 Perspective common field search means 3, 4 fields 5 Hypothesis definition means 6 Database 7 Hypothesis verification means 8 Revision proposal

Claims (7)

On the computer,
Refers to a classification system information storage means for storing classification system information indicating a classification system in which a classification viewpoint is defined in each of a plurality of fields and a category for classification according to the defined viewpoint is provided. Extract at least two fields with categories,
Compare the categories established in the common viewpoint of each extracted field,
Define a hypothesis that a category that exists only in some of the extracted fields is an additional target category, and the additional target category is to be added to another field,
Referring to a database in which information to be classified by the classification system is stored, and when the additional target category is added according to the hypothesis, additional category conforming information that belongs to the additional target category is detected,
Determining the suitability of the hypothesis based on the amount of additional category conformance information detected;
Outputting the hypothesis determined to be appropriate as a revised draft,
Classification system revision support program characterized by having processing executed.   When a plurality of combinations of at least two fields in which categories are provided from a common viewpoint are extracted, a combination having a larger number of categories having the same definition among the categories provided from the common viewpoints of each field, 2. The classification system revision support program according to claim 1, wherein the priority of hypothesis setting processing is increased. When detecting the additional category conformity information,
Among the extracted fields, a field that already has the additional target category as a category according to the common viewpoint is set as a reference field, and a field that does not have the additional target category as a category according to the common viewpoint As the revision target field, the information belonging to the reference field in the database is analyzed to extract the additional target category feature information indicating the characteristics of the information belonging to the additional target category, and each of the revision target fields in the database It is determined whether each piece of information belonging to the revision target field corresponds to the additional category conforming information depending on whether the information includes the additional target category feature information.
The classification system revision support program according to claim 1, wherein the program is executed. When extracting the additional target category feature information, the computer,
Information that belongs to the additional target category by comparing a keyword obtained from information belonging to the additional target category among information belonging to the reference field in the database and a keyword obtained from information not belonging to the additional target category A feature keyword indicating a feature of the feature, and the feature keyword as the addition target category feature information,
4. The classification system revision support program according to claim 3, wherein the program is executed. When extracting the additional target category feature information, the computer,
A classifier obtained by performing machine learning using information belonging to the additional target category as a positive example and information not belonging to the additional target category as negative examples among information belonging to the reference field in the database, and obtained as a result of machine learning The additional target category feature information
4. The classification system revision support program according to claim 3, wherein the program is executed. Refers to a classification system information storage means for storing classification system information indicating a classification system in which a classification viewpoint is defined in each of a plurality of fields and a category for classification according to the defined viewpoint is provided. A viewpoint common field search means for extracting at least two fields in which categories are provided;
Compare the categories provided from the common viewpoint of each extracted field, and select categories that exist only in some of the extracted fields as additional target categories, and set the additional target categories as other fields. A hypothesis definition means for defining a hypothesis to be added to
Referring to a database in which information to be classified by the classification system is stored, when the additional target category is added according to the hypothesis, additional category conforming information belonging to the additional target category is detected and detected Hypothesis verification means for determining the suitability of the hypothesis based on the amount of the additional category conforming information that has been determined, and outputting the hypothesis determined to be appropriate as a revision plan;
A classification system revision support device characterized by comprising: Computer
Refers to a classification system information storage means for storing classification system information indicating a classification system in which a classification viewpoint is defined in each of a plurality of fields and a category for classification according to the defined viewpoint is provided. Extract at least two fields with categories,
Compare the categories established in the common viewpoint of each extracted field,
Define a hypothesis that a category that exists only in some of the extracted fields is an additional target category, and the additional target category is to be added to another field,
Referring to a database in which information to be classified by the classification system is stored, and when the additional target category is added according to the hypothesis, additional category conforming information that belongs to the additional target category is detected,
Determining the suitability of the hypothesis based on the amount of additional category conformance information detected;
Outputting the hypothesis determined to be appropriate as a revised draft,
Classification system revision support method characterized by this.

Images