JP2003157262A - Patent retrieval device, control method therefor, control program and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To clearly present referring and to-be-referred relation among a plurality of patent documents and to extract the patent document of approximate technical contents though there is not the referring relation. SOLUTION: The retrieval processing part 102 of a patent retrieval system 100 analyzes a retrieving source patent document and extracts a word or a composite word included in the patent document for setting conditions as a retrieval condition word/phrase. Then, by referring to a database part 11, a word/phrase for retrieval and retrieval condition word/phrase are compared with each other to retrieve a patent document similar to the patent document of the retrieval source as a patent document highly related with the patent document of the retrieval source.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a patent search device, a control method for the patent search device, a control program, and a recording medium, and more particularly to a relationship between a plurality of patent documents or a relationship between claims in a patent document. A technique for clearly presenting to a searcher.

[0002]

2. Description of the Related Art With the development of technology in recent years, the number of patent applications has become enormous. Along with this, the volume of patent documents is also enormous, preventing duplicate research, checking for infringement of rights, investigating before application, grasping technology development trends of other companies, determining the direction of research and development, etc. In fact, it is not easy to search for patent documents. A patent search system has been conventionally known as a means for reducing the load in such patent document search. In the conventional patent search system, one or more keywords that are considered to be included in the patent document to be searched are input, the patent document including the keyword or keyword group is extracted by referring to the database, and the extracted patent document It was configured to generate a list and display or output it.

[0003]

In the above-mentioned conventional patent search system, although a "leakage-free" search can be performed, if a keyword is selected incorrectly, a large amount of unnecessary information is included. There was a problem that it would end up.
And in such a case, there was a problem that the time for the investigation itself would be enormous. In addition, there is a problem in that if the number of keywords is unnecessarily increased and narrowed down, related patent documents will also be omitted. Further, even when a plurality of patent documents are extracted, there is a problem that the mutual relationship between the respective patent documents cannot be understood unless the contents of the documents are read in detail. In other words, even if a plurality of related patent documents are extracted, it is necessary to refer to information described in the plurality of documents to some extent to determine which patent document is closer to the basic patent. Therefore, an object of the present invention is to provide a patent search device capable of extracting a patent document having technical contents that are similar to each other even if there is no reference relationship, by clearly presenting references between a plurality of patent documents and referenced relationships. It is intended to provide a control method, a control program, and a recording medium of a patent search device.

[0004]

In order to solve the above problems, a patent search device stores a word or a compound word included in a patent document to be searched as a search term phrase in advance in association with the patent document. Database department,
The search source patent document is analyzed to compare the search term phrase and the search condition phrase with a phrase extraction unit that extracts a word or a compound word included in the condition setting patent document as a search condition phrase, and the search source patent document And a search unit for searching a patent document highly similar to the document as a patent document highly related to the search source patent document. According to the above configuration, the word extraction unit of the patent search device is
The search source patent document is analyzed, and a word or a compound word included in the condition setting patent document is extracted as a search condition phrase. The search unit refers to the patent search database unit to compare the search term phrase and the search condition phrase, and a patent document having a high similarity to the search source patent document is highly relevant to the search source patent document. Search as a patent document.

In this case, a reference destination extraction unit for extracting a reference relationship between the search source patent document and the highly related patent document may be provided based on the description in each patent document. . Further, the patent search database unit may include a reference destination database unit that stores information about patent documents having a reference relationship in association with the search source patent document based on the extracted reference relationship. . Further, based on the information of the reference destination database unit, a chart creation unit that creates a chart representing a reference relationship between the search source patent document and the highly related patent document,
A display unit that presents the chart created by the chart creating unit may be provided.

Furthermore, a morphological analysis unit for performing morphological analysis of the search source patent document to extract the word or the compound word, and an importance of the extracted word or compound word in the search source patent document. An importance degree calculating unit for calculating and a registration unit for registering the extracted word or the compound word in the patent search database unit as the search term phrase in association with the importance degree may be provided. Further, the search unit may search for patent documents having high relevance based on a distance between a vector corresponding to the search term phrase and a vector corresponding to the search condition phrase.

Further, the above-mentioned patent document has a plurality of "claims".
The search unit may include a claim analysis unit that extracts a dependency relationship between the plurality of “claims” based on the search term phrase. Furthermore,
The claim analysis unit is based on the description of another “claim” included in the “claim” to be searched and the names of the inventions of the “claim” and other “claims” to be searched. You may make it extract a subordinate relationship. Further, the claim analysis unit has the description of the other "claim" in the "claim" of the search target, and the "claim" of the search target.
It may be determined that the subordinate relationship exists when the title of the invention of 1) and the title of the invention of the other claim are the same. Further, the patent document includes a description of “claim”, and the search unit includes a claim analysis unit that extracts a constituent element included in the description of “claim” based on the search term phrase. May be.

Further, in a control method of a patent search apparatus having a patent search database section for storing a word or a compound word included in a patent document to be searched as a search term phrase in advance in association with the patent document, Analyzing, the phrase extraction process of extracting a word or a compound word included in the condition setting patent document as a search condition phrase, comparing the search term phrase and the search condition phrase,
A search process for searching a patent document having a high similarity to the search source patent document as a patent document highly related to the search source patent document. In this case, a reference destination extraction process for extracting a reference relationship between the search source patent document and the highly related patent document may be provided based on the description in each patent document.

The patent search database section includes a reference destination database section that stores information about patent documents having a reference relationship in association with the search source patent document based on the extracted reference relationship. Based on the information in the reference destination database unit, a chart creation process of creating a chart representing a reference relationship between the search source patent document and the highly related patent document, and a chart created by the diagram creation unit. And a display process to present. Further, a morphological analysis process of extracting the word or the compound word by performing a morphological analysis of the search source patent document, and an importance calculation for calculating the importance of the extracted word or compound word in the search source patent document. A process and a registration process of associating the extracted word or the compound word with the importance degree and registering the word or the compound word in the patent search database unit as the search term phrase may be provided.

Further, the searching step may search for patent documents having high relevance based on the distance between the vector corresponding to the search term phrase and the vector corresponding to the search condition phrase. Further, the patent document includes descriptions of a plurality of “claims”, and the search process extracts a subordinate relationship between the plurality of “claims” based on the search term phrase. May be provided. Further, the claim analysis process is based on the description of another "claim" included in the "claim" to be searched and the names of the inventions of the "claim" and other "claims" to be searched. You may make it extract the said subordinate relationship.

Furthermore, in the claim analysis process, there is a description of the other "claim" in the "claim" of the search target, and the invention name of the "claim" of the search target and the It may be determined that there is the subordinate relationship when the names of the inventions in other "claims" are the same. Also,
The patent document includes a description of “claim”, and the search process includes a claim analysis process of extracting a constituent element included in the description of “claim” based on the search term phrase. Good.

Further, a control program that causes a computer having a patent search database unit that stores in advance a word or a compound word included in a patent document to be searched as a search term phrase in association with the patent document as a patent search device is: The search source patent document is analyzed, and a word or a compound word included in the condition setting patent document is extracted as a search condition phrase, and the search term phrase and the search condition phrase are compared, with respect to the search source patent document. It is characterized in that a patent document having a high similarity is searched as a patent document having a high relevance to the search source patent document. In this case, the reference relationship between the search source patent document and the highly related patent document may be extracted based on the description in each patent document. Moreover, the said patent document may contain description of several "claim", and you may make it extract the subordination relationship between these "claim" based on the said search term phrase. Further, the patent document,
It is also possible to include the description of “claim” and extract the constituent elements included in the description of “claim” based on the search term phrase. Further, each of the above control programs may be recorded in a recording medium.

[0013]

BEST MODE FOR CARRYING OUT THE INVENTION Next, preferred embodiments of the present invention will be described with reference to the drawings. The present embodiment is a case where the present invention is applied to a patent search system. [1] Patent Search Database Registration System First, a patent search database registration system for constructing a patent search database used in the patent search system will be described. [1.1] Configuration of Patent Search Database Registration System FIG. 1 is a schematic block diagram of the patent search database registration system. Patent search database registration system 1
0 roughly includes a database unit 11 that accumulates various data as a database, and a registration processing unit 12 that registers various data in the database unit 11. Here, the patent search database registration system 10
Can be realized in a computer system, and the function of the registration processing unit 12 is realized by a program that can be executed by a microprocessor and corresponds to each unit that constitutes the registration processing unit 12. Further, such a program may be directly executed from a recording medium such as a semiconductor memory or a CD-ROM. It is also possible to install the program in an external storage device in advance and execute it. Furthermore, the program may be installed each time it is executed prior to execution or only once at the beginning via a network such as the Internet.

[1.1.1] Configuration of Database Unit First, the configuration of the database unit will be described. The database unit 11 is built in an external storage device such as a hard disk. The database unit 11 is roughly classified into a word database 15, a constituent requirement (invention specifying matter) database 16, and a dependency relation database 17.
And a reference document database 18. The word database 15 is a database for storing a word or a compound word included in the patent document PD in association with a predetermined specific classification (field).

FIG. 2 shows the basic structure of the word database. The word database 15 roughly includes a page table 21, a keyword table 22, and a word table 23. In FIG. 2, one page table 21 is used for one keyword table 2
2 shows only the word table 23 and the word table 23, in reality, a plurality of systems of the keyword table 22 and the word table 23 are configured for one page table 21 to form a tree structure (tree structure) as a whole. ing. The page table 21 includes page ID data 31, identification number data 32, title data 33, update date / time data 34, bibliographic item data 35, and link count data 3
6 is provided. Note that this data structure is an example,
It is not limited to this. For example, data such as the number of pages and the technical field of the corresponding publication may be incorporated.

The page ID data 31 is data uniquely assigned to each patent document (patent document) PD to be registered, and is data for specifying each patent document PD. The identification number data 32 is data having a value assigned to each patent document PD by the patent office, such as a publication number. The title data 33 stores a name indicating the content of each patent document PD. For example, when the patent document PD is a patent publication, it is the title of the invention. The update date / time data 34 stores the registration date / time of each patent document PD. If the patent document PD is a patent gazette, the bibliographical matter data 35 includes the country of publication of the gazette, the gazette type, the publication number, the publication date, the international patent classification information (FI classification number, etc.), the presence or absence of an examination request,
Number of claims, application form, total number of pages, application number, application date, application number of basic application claiming priority, application number of basic application claiming priority, application date of basic application claiming priority, applicant information (identification Number, name or name, address or whereabouts), inventor information (name, address or whereabouts), agent information (identification number,
Name, qualification), theme code, F term, etc. That is, the bibliographic item data 35 is data for storing various bibliographical information relating to each patent document PD. The linked number data 36 is data for storing the number of documents of other patent documents to which each patent document PD is referred.

The keyword table 22 is a table for associating the page table 21 and the word table 23 with each other. And the keyword table 22
Includes page ID data 41, word ID data 42, appearance frequency (cost) data 43, importance level data 44, and paragraph flag data 45. Page ID data 4
1 is data stored in the data corresponding to the page ID data 31 of the page table 21. That is, if the same data as the page ID data 31 of the page table 21 is stored, the keyword table 22 is
This indicates that data related to the page table 21 having the same page ID data is stored. The word ID data 42 is data for associating with the word table 23 that stores information about corresponding words.

The number of appearances (cost) data 43 is the word I
Page ID data 41 of the word specified by the D data 42
It is data corresponding to the number of appearances in the patent document specified by. That is, it is data indicating how many times a certain word is used in the patent document. Importance data 44
Stores the TFIDF value calculated by the TFIDF method as the importance of the word specified by the word ID data 42 in the patent document specified by the page ID data 41. This TFIDF value is calculated based on IDF data (IDF value; corresponding to the number of documents in which the corresponding word appears in all patent documents) and the number-of-occurrence data 43 stored in a word table described later. That is,
The smaller the number of documents in which a certain word appears and the larger the number of occurrences of the word in each document, the more important the word is to be recognized.

The paragraph flag data 45 is data indicating in which paragraph in the patent document specified by the page ID data 41 the word specified by the word ID data 42 is detected. In this case, when the patent document is an ordinary document such as a paper or a book, the paragraph includes not only the meaning of general paragraphs but also the concept of chapters and sections. Further, when the patent document is a patent gazette or the like, paragraphs of the title column of the invention, the claims column, the detailed explanation column of the invention, the brief explanation column of the drawing, the abstract column, etc. Handling.

The word table 23 includes word data 51,
The word ID data 52 and the IDF data 53 are provided. In the word data 51, the words after morphological analysis are registered. The word ID data 52 is data for identifying a word registered as the word data 51. The IDF data 53 is registered with IDF values corresponding to the number of documents in which the word registered as the word data 51 appears in all patent documents. The constituent requirement database 16 stores data of constituent requirements (so-called invention specifying matters) included in the claims from patent publications, patent publications, and the like, which are patent documents. Specifically, the name of each component, a word that characterizes the action of each component, and the like are stored. The dependency database 17 stores data relating to a dependency which is a connection between claims.

FIG. 3 shows the data format of the dependency database 17. The dependency database 17 includes page ID data 55, cited claim data 56, and cited claim data 57. The page ID data 55 is the page ID data 31 of the page table 21.
And the cited claim data 56 and the cited claim data 57 corresponding to the page ID data 55.
Represents the subordinate relationship of the claims included in the patent document corresponding to the page ID data 55. In the cited claim data 56, a claim number of a claim cited by another claim is registered. In the cited claim data 57, the claim number of a claim that refers to another claim is registered.

Specifically, in the description of the claims of the patent document specified by page data = 1,
And claim 2 are independent claims, claim 3 is a dependent claim quoting claim 1 and claim 2, and description of claims of patent document specified by page data = 2 In claim 1, when claim 1 is an independent claim and claim 2 is a dependent claim that refers to claim 1, data as shown in FIG. 4 is registered in the dependent relationship database 17. It That is, as shown in FIG. 4, the following data items are registered. Page ID data = 1, Cited claim data = 1, Cited claim data = 3 Page ID data = 1, Cited claim data = 2, Cited claim data = 3 Page ID data = 2, Cited claim Data = 1, cited claim data = 2

The reference document database 18 stores, for each patent document, data relating to the reference document referred to in the document or the reference document referred to in the examination or the like. Specifically, the reference document number (patent publication number, patent number, ...), Document name, Document reference page, Document issuer, Author name, etc. are stored as necessary.
As shown in FIG. 8, the reference document database 18 includes reference destination page ID data 81 and reference source page ID data 8
2, title data 83, document type data 84, and document bibliographic information data 85. Reference page ID
The data 81 is page ID data for identifying the patent document referred to. The reference page ID data 82 is
This is page ID data for specifying the patent document of the reference source.

The title data 83 registers the reference title of the reference. In the case of the above example, "Japanese Patent Laid-Open No. -12"
"345 gazette" is the title of the document. The document type data 84 is data representing the type of the patent document referred to. For example, when the type of the referred patent document is a domestic patent publication, the document classification data = 0, when it is a foreign patent publication, the document classification data = 1, and when it is a paper, the document classification data = 2. Is set. Bibliographic information data 85
Stores bibliographic information such as a document name, a document reference page, a document issuer, and an author name in a predetermined order in a text data format such as a CSV format. The bibliographic information data 85 need not be provided unless these pieces of information are needed.

[1.1.2] Configuration of Search Processing Unit Next, the configuration of the search processing unit 12 will be described with reference to FIG. 1 again. The search processing unit 12 is roughly classified into a morpheme analysis unit 61, an importance degree calculation unit 62, a claim analysis unit 63, and
A morphological analysis dictionary unit 64, a reference document extraction unit 65, a standardization unit 66, and a stop word processing unit 67 are provided. The morpheme analysis unit 61 registers in the word database 15 nouns, sahen nouns, other nouns or words similar to sahen nouns that frequently appear by morphologically analyzing the patent document PD. Prior to the registration of words in the word database 15 by the morphological analysis unit 61, the standardization unit 66 standardizes (unifies) the words. For example, the word “PC”,
When three types of words "personal computer" and "personal computer" are obtained by morphological analysis, these words have the same meaning and are therefore unified into the word "personal computer". In addition, the stop word processing unit 67
In the patent field, processing is performed so that words that frequently appear regardless of the type of patent document are not registered in the word database 15. For example, words such as “claim”, “application”, “patent”, “present invention” and “utility model” are not registered in the word database 15.

The importance calculating unit 62 calculates the importance of the word registered by the morphological analysis using the TFIDF method.
It is calculated as a FIDF value and registered in the word database 15. The claim analysis unit 63 roughly includes a dependency relationship analysis unit 63A and a constituent requirement extraction unit 63B.
When the patent document is a patent gazette or a utility model gazette, the dependency relationship analyzing unit 63A analyzes the claims and the utility model registration claim, and examines the dependency relationship between the claims. Further, the constituent requirement extracting unit 63B extracts constituent requirements (for example, an invention specifying matter) from the claims and the utility model registration request. Morphological analysis dictionary unit 6
Reference numeral 4 stores a morpheme analysis dictionary of a generally known form used when the morpheme analysis unit 61 performs morpheme analysis. For each patent document, the reference document extracting unit 65 extracts the reference document referred to in the document or the reference document referred to in the examination or the like, and registers it in the reference document database 18.

[1.2] Operation of Patent Search Database Registration System Next, the operation of the patent search database registration system will be described. In the following description, the case of registering a patent publication as a patent document will be described as an example. FIG. 5 shows a flowchart of the patent search database registration processing. First, the patent search database registration system 10 performs morphological analysis (step S1) and registers it in the word database (step S2). Specifically, the morphological analysis unit 61
When performing morphological analysis on a patent document, a unique number for each patent document is assigned as a page ID based on a predetermined condition and registered in the page ID data of the page table and the page ID data of the keyword table. Next, the morpheme analysis unit 61 performs morpheme analysis of patent publications that are patent documents, and registers the publication number as the identification number data 32.

Then, the morphological analysis unit 61 extracts the title of the invention and registers the title of the invention as title data 33. Then, the morpheme analysis unit 61 registers the registration date and time in the database registration system for the patent publication in the update date and time data. Also, the morphological analysis unit 6
1 indicates from the patent gazette, publication country, gazette type, publication number, publication date, international patent classification information, presence or absence of examination request, number of claims, application form, total number of pages, application number, application date, priority The application number of the claimed basic application, the application number of the claimed basic application, the application date of the claimed basic application, the applicant information, the inventor information, the agent information, the theme code, the F-term, etc. are extracted, and the bibliographic data Register as 35. Next, the patent search database registration system calculates the importance of each word based on the result of the morphological analysis (step S3) and registers it in the keyword table 22 of the word database 15 (step S2).

In this case, the importance calculator 62
The importance level stored as the importance level data is calculated based on the appearance frequency (cost) data 43 of the keyword table and the IDF data 53 of the word table. Specifically, the number of appearances of each word (keyword) included in all registered patent documents is TF, the number of all registered patent documents is N, and the importance calculation target among all registered patent documents is When the number of patent documents including a word is n, the TFIDF value is calculated as importance data by the following formula. IDF = log (N / n) TFIDF = TF · IDF In this case, the calculated IDF is registered in the word table as IDF data. Next, the patent search database registration system analyzes the dependency relation between claims which is a part of the analysis of the claims (step S4) and registers it in the dependency relation database 17 (step S5).

In this case, the dependency analysis unit 63A
Refers to the description such as "in the □□ method described in claim XX" in the claims and the utility model registration claim, "... ■ device in claim △△ characterized by ...." Based on this, the dependency relationship between each claim is analyzed, and the dependency relationship as shown in FIG. 4 is registered in the dependency relationship database. Next, the patent search database registration system extracts the constituent requirements that are part of the claim analysis (step S).
6). In this case, the constituent requirement extracting unit 63B extracts constituent requirements (invention specifying matters) from the description of the claims and registers them in the constituent requirement database 16 (step S7).
Specifically, a case where the constituent requirements are extracted from the following claims will be described.

[Claim 1] n ordered in time series
A method of associating a plurality of documents, calculating a similarity between the n documents, creating a similarity matrix from the similarity using a time constraint, and calculating the similarity matrix from the documents. And a step of converting into an adjacency matrix showing the association of the document.

In this case, the following three constituent elements are extracted by the keyword "step". "Step of calculating the similarity between the n documents""Step of creating a similarity matrix from the similarity using a time constraint", "Conversion of the similarity matrix into an adjacency matrix showing association of the documents" Steps to be performed ”and the extracted configuration requirements are stored in the configuration requirement database 1
6 will be registered.

Specifically, the configuration requirement database 16
6, includes page ID data 71, complaint ID data 72, and constituent text data 73. The page ID data 71 is data similar to the page ID data 31 of the page table 21, and the complaint ID data 72 corresponding to the page ID data 71.
And the configuration requirement text data 73 is the page ID
The content of the claim included in the patent document corresponding to the data 71 is represented. The claim ID data 72 is data for specifying the claim number of the text data stored in the constituent requirement text data 73.

The constituent requirement text data 73 is data for storing the constituent requirement as text data.
Then, when the above claims are registered in the constituent requirement database 16, it becomes as shown in FIG. That is, the number = 231 for identifying the patent document is stored as the page ID data 71, the number = 1 representing claim 1 is stored as the claim ID data 72, and the text data corresponding to the content of claim 1 = “ The step of calculating the similarity between the n documents ”is stored as the constituent requirement text data 73.

Next, the patent search database registration system 1
0 extracts a reference document (step S8). In this case, the reference document extraction unit 65 is a patent document (patent document).
Within one of the
It is detected whether or not a citation (such as a paper) is cited, and information on the reference destination (publication number, title, article title, etc.) is registered in the reference database. Further, when the referred patent document is already registered, the value of the linked number data 36 of the page table 21 corresponding to the referred patent document is updated.

Specifically, in the patent gazette which is a patent document, see "according to Japanese Patent Laid-Open No. 12345 -..." and "Document" Fast Algorithms for Mining Association "...
The description such as “...” is extracted and registered in the reference document database 18.

[1.3] Configuration of Patent Retrieval System FIG. 9 is a schematic block diagram of the patent retrieval system.
9, the same parts as those in the patent search database registration system of FIG. 1 are designated by the same reference numerals, and detailed description thereof will be omitted. The patent search system 100 is roughly divided into
A database unit 11 that stores various data as a database, a search processing unit 102 that searches for a patent using the database unit 11, a display unit 103, and an input unit 104 are provided.

Here, the patent search system 100 can be realized in a computer system, and the function of the search processing unit 102 can be realized by a program executable by a microprocessor corresponding to each unit constituting the search processing unit 102. It Further, such a program may be directly executed from a recording medium such as a semiconductor memory or a CD-ROM. It is also possible to install the program in an external storage device in advance and execute it. Furthermore, the program may be installed each time it is executed prior to execution or only once at the beginning via a network such as the Internet. Search processing unit 102
Is the search source patent document input through the input unit 104 (or specific information for specifying the search source patent document)
Based on the above, a similar search, a claim search, or a reference destination search described later is performed. The display unit 103 displays various data including the search result and presents it to the search operator. The input unit 104 includes a keyboard, a mouse, a tablet, a scanner, a removable disk device, a communication interface unit, and the like, and inputs data corresponding to the search source patent document and sets various search conditions.

[1.3.1] Configuration of Search Processing Unit Next, the configuration of the search processing unit will be described. Search processing unit 1
02 is roughly classified into a syntactic / morphological analysis unit 111.
And a morphological analysis dictionary unit 112 and a chart creation unit 113.
A real search processing unit 114, a standardization unit 115, and a stop word processing unit 116. The syntax analysis / morpheme analysis unit 111 functions as a phrase extraction unit, and includes a morpheme analysis unit 121, a claim analysis unit 122, and a reference document extraction unit 123. The morphological analysis unit 121 uses the morphological analysis dictionary 64 to morphologically analyze the patent document and extracts frequently occurring nouns, sahen nouns, other nouns, or words similar to sahen nouns.

When the morphological analysis unit 121 outputs the morphological analysis result, the standardization unit 115 standardizes (unifies) the words. In addition, the stop word processing unit 116
Performs processing so that a word that frequently appears in the patent field regardless of the type of patent document is not output as a morphological analysis result. When the search source patent document is a patent publication, a utility model publication, or the like, the claim analysis unit 122
Analyze claims and utility model registration claims, etc., and examine the subordination between claims. In addition, constituent elements (for example, invention specifying matters) are extracted from the claims and the utility model registration claims. Reference destination extraction unit 123
For the search source patent document, the reference document referenced in the search source patent document or the reference document referred to in the examination is extracted. Morphological analysis dictionary unit 112
Stores a morphological analysis dictionary used when the morphological analysis unit 121 performs morphological analysis.

The chart creation unit 113 charts the search result and presents it to the search operator. Real search processing unit 11
4 functions as a search unit, and if roughly classified, the similar search unit 1
31, a claim search unit 132, and a reference destination search unit 133
And are equipped with. The similarity search unit 131 searches for patent documents similar to the search source patent document, extracts patent documents having a similar relationship, and grasps the similarity relationship. Claim search section 1
32 grasps the subordinate relationship of the claims (claims) of the search source patent document. The reference document search unit 133 extracts the patent document referred to in the search source patent document or the patent document referred to in the examination or the like.

[1.4] Operation of Patent Search System Next, the operation of the patent search system will be described. In the following description, the case where a patent publication is used as a search source patent document will be mainly described. FIG. 10 shows a flowchart of patent search processing. In this case, prior to the patent search process, the operator performs a similar search process in which similar patent documents are searched, or a claim search process in which the relationship between claims of the patent document is examined. Is specified, the similarity search flag is set when the similar search process is performed, and the claim flag is set when the complaint search process is performed. First, the patent search system 100 determines whether the search source patent document is not registered in the database unit 11 (step S11).

In the determination in step S11, if the search source patent document is already registered in the database (step S11; No), it is determined whether the similarity search flag or the claim flag is set (step S11). S12). In the determination of step S12,
When the similarity search flag is set (step S12; similarity flag set), the reference destination search unit 133 of the actual search processing unit 114 searches the reference document database 18 and refers to the search source patent document. The patent documents are extracted (step S13). Next, the similarity search unit 131 of the actual search processing unit 114 reads out the words included in the search source patent document by referring to the word database 15 and performs a similarity search for each word (step S14), and searches Pattern matching is performed for distance calculation with other patent documents registered in the word database 15 and constituent elements of claims.

More specifically, when the search source patent document is a patent publication in Japan, for example, each word included in the detailed description and abstract of the invention is expressed as a vector and registered in the word database. The distance calculation processing is performed with respect to the vector corresponding to the word included in another patent document. In addition, pattern matching processing is performed between the constituent features of the claims that form the claims in the search source patent document and the constituent features registered in the constituent feature database, and the degree of similarity is obtained. The patent documents within a predetermined similarity range are specified based on the degree of similarity obtained by the distance calculation process and the pattern matching process.
Further, the same distance calculation processing and pattern matching processing are also performed for the patent documents referred to by the search source patent document. Then, when the similarity search process ends, the chart creation unit 113 performs a process for tabulating the similarity search results (step S15). When the chart creating unit 113 completes the processing for tabulating the similar search results, the display unit 103 displays the chart corresponding to the similar search results (step S16).

FIG. 11 shows an example of a chart corresponding to the similarity retrieval result. In the following description, the expression "patent publication" includes not only patent publications in Japan but also published patent publications and other publications. In FIG. 11, patent publication X is the search source patent literature. Patent publication X
Patent documents directly or indirectly linked to
It is a patent document referred to directly or indirectly.
Further, patent documents that are not tied to any of the patent documents,
It is a patent document that is a similar document but not in a direct reference relationship. In this case, the closer the two-dimensional distance between patent documents, the higher the degree of similarity. Further, the patent documents displayed on the left side of the display position of the patent publication X and directly linked to the patent publication X (US patent publication S, patent publication C, patent publication B, patent publication D, university paper R) Is a patent document directly referenced by Patent Publication X. That is, US patent publication S, patent publication C, patent publication B, patent publication D, and university thesis R are patent publication X, which is the search source patent literature.
Patent documents directly described and referenced therein.

Further, these US patent publication S, patent publication C, patent publication B, patent publication D, university thesis R and patent publication X
The closer the two-dimensional distance between and is, the higher the similarity is. Further, the numbers displayed in the frames representing the respective patent documents are the number of patent documents referring to the patent documents. In the case of the example of this figure, since the patent publications A and E have a large number of references, it can be determined that the possibility of a basic patent is high. Therefore, when the patent publication X is a patent publication, the patent applicant is
It is considered that the invention described in Patent Publication E should be referred to first. Furthermore, since there are patent publications N, Q, and P that are similar to patent publication X, it is understood that the inventions described in these patent publications also require attention. In this case, if any of the patent documents is selected on the display screen of the display unit 103 and the search type is designated, the search process is similarly performed with the selected patent document as the search source patent document. With this configuration, it is possible to continue searching hierarchically,
A desired patent document can be efficiently searched.

Further, when a plurality of patent documents are connected by a line segment, the line segment is color-coded for systematic classification, for each applicant,
It is possible to make it easier to understand the relationship between patent documents by changing the color in the frame representing the patent documents or displaying blinking (blinking) for each inventor and agent (patent attorney, etc.). Becomes On the other hand, when the claim flag is set in the determination of step S12 (step S12; claim flag set), the actual search processing unit 11
The claim retrieval unit 132 of No. 4 retrieves the dependency relation between claims by searching the constituent requirement database and the dependency relation database (step S17). Then, when the complaint search processing is completed, the chart creation unit 113 performs processing for charting the complaint search results (step S1).
5). When the chart creating unit 113 completes the processing for tabulating the complaint search results, the display unit 10
A chart corresponding to the complaint search result is displayed in step 3 (step S16).

FIG. 12 shows an example of a chart corresponding to the complaint search result. In this case, each claim shall have the following description. Claim 1: "Knowledge extraction method characterized by including ..." Claim 2: "In the knowledge extraction method according to Claim 1,
...... ”Claim 3:“ In the knowledge extraction method according to Claim 2,
...... "Claim 4:" Knowledge extraction device characterized by including ... "Claim 5:" In the knowledge extraction device according to claim 4,
...... ”Claim 6:“ In the knowledge extraction device according to claim 4,
...... ”Claim 7:“ In the knowledge extraction device according to claim 4,
...... ”Claim 8:“ In the knowledge extraction device according to claim 4,
...... ”Claim 9:“ Knowledge extraction program characterized by having …… ”

The claim search unit 132 detects that claims 1, 4, and 9 are independent claims based on the description of each claim. That is, a claim that does not include the description of another claim in the description of the claim or a claim that includes only the description of another claim with a different title of the invention is regarded as an independent claim. Further, the claim search unit 132 has the same name as the invention of the claim described before the claim, and the description of another claim (for example, “Claim XX”). It is judged that a claim including ") is a dependent claim dependent on the other claim. That is, it is detected that claim 2 is a dependent claim of claim 1, claim 3 is a dependent claim of claim 2, and claims 5 to 8 are dependent claims of claim 4, respectively. As a result, as shown in FIG. 12, the frame corresponding to claim 1 and the frame corresponding to claim 2 are connected by a line segment, and the frame corresponding to claim 2 and the frame corresponding to claim 3 are lined. Connected and displayed by minutes. Further, in the frame corresponding to claim 4, the frames corresponding to claims 5 to 8 are displayed in parallel connected by line segments. Further, the frame corresponding to claim 9 is displayed independently. With such a display, the searcher can intuitively grasp the independent claim specifying the basic content in the claim group and easily establish the subordination relationship between the claims. I can figure it out.

[1.5] Effects of the Embodiments As described above, according to the present embodiment, when performing a patent search, an operator conducts a "least leak" survey without inputting keywords. In addition, it is possible to suppress the inclusion of unnecessary information. Further, it is possible to clearly present references and referenced relationships among a plurality of patent documents, and to extract patent documents having similar technical contents even if there is no reference relationship. Furthermore, it is possible to clarify the dependent relationship of claims in the same patent document and present them to the operator in a state where they can be easily visually grasped.

[2] Modification of Embodiment [2.1] First Modification In the above description, the database unit 11 is integrated with the registration processing unit 12 (or the search processing unit 102). Both can be configured as a distributed processing system via a network. In this case,
Further, each database 1 that constitutes the database unit 11
5, 16, 17, 18 and the morphological analysis dictionary 64 are configured to be stored in another database server via a network, and are used by a computer system functioning as a plurality of registration processing units 12 (or search processing units 102). A possible configuration is also possible.

[2.2] Second Modification In the above description, the standardization unit 66 (or the standardization unit 115) has been described as an indispensable configuration, but the capacity of the database is somewhat small even if the standardization unit 23 is not necessarily provided. Although it has the disadvantage of increasing, it is possible to obtain almost the same effect.

[0053]

According to the present invention, when conducting a patent search, it is possible to carry out a "least leaked" search and suppress the inclusion of unnecessary information. Further, it is possible to clearly present references and referenced relationships among a plurality of patent documents, and to extract patent documents having similar technical contents even if there is no reference relationship. Furthermore, the dependency relationship between claims in the same patent document can be clarified and presented.

[Brief description of drawings]

FIG. 1 is a schematic block diagram of a patent search database registration system.

FIG. 2 is a basic configuration diagram of a word database.

FIG. 3 is an explanatory diagram of a data format of a dependency database.

FIG. 4 is an explanatory diagram of a data registration example of a dependency database.

FIG. 5 is a flowchart of a patent search database registration process.

FIG. 6 is an explanatory diagram of a data format of a constituent requirement database.

FIG. 7 is an explanatory diagram of a data registration example of a constituent requirement database.

FIG. 8 is an explanatory diagram of a data format of a reference document database.

FIG. 9 is a schematic block diagram of a patent search system.

FIG. 10 is a flowchart of patent search processing.

FIG. 11 is an explanatory diagram of an example of a chart corresponding to a similarity search result.

FIG. 12 is an explanatory diagram of an example of a chart corresponding to a complaint search result.

[Explanation of symbols]

10 ... Patent search database registration system 11 ... Database section 12 ... Registration processing unit 15 ... Word database 16: Database of constituent requirements (items specifying invention) 17 ... Dependency database 18 ... Reference database 21 ... Page table 22 ... Keyword table 23 ... Word table 31 ... Page ID data 32: Identification number data 33 …… Title data 34: Update date / time data 35 …… Bibliographic data 36 ... Linked data 100 ... Patent search system 102 ... Search processing unit 103 ... Display section 104 ... Input section 111 ... Syntax analysis / morphological analysis unit 112 ... Dictionary for morphological analysis 113 ... Chart creation department 114 ... actual search processing unit 115 …… Standardization Department 116 ... Stop word processing section 121 ... Morphological analysis unit 122 ... Claim analysis unit 123 ... Reference document extraction unit

Claims (24)

[Claims] 1. A patent search database unit that stores a word or a compound word included in a patent document to be searched as a search term phrase in advance in association with the patent document, and analyzes the search source patent document to set the condition. A word extraction unit that extracts a word or a compound word included in a patent document as a search condition phrase, compares the search term phrase and the search condition phrase, and obtains a patent document having high similarity to the search source patent document. A patent search device, comprising: a search unit that searches a patent document highly relevant to a search source patent document; 2. The patent search device according to claim 1, wherein a reference destination extraction for extracting a reference relationship between the search source patent document and the highly related patent document based on the description in each patent document. Patent search device characterized by comprising a section. 3. The patent search device according to claim 2, wherein the patent search database section associates information about patent documents having a reference relationship with the search source patent document based on the extracted reference relationship. A patent search device comprising a reference destination database unit for storing. 4. The patent search device according to claim 3, wherein a chart representing a reference relationship between the search source patent document and the highly related patent document is created based on the information in the reference destination database section. And a display section for presenting the chart created by the chart creating section. 5. The patent search device according to claim 1, further comprising: a morphological analysis unit that performs morphological analysis of the search source patent document to extract the word or the compound word. And an importance degree calculation unit that calculates the degree of importance of the word or the compound word in the search source patent document, and the extracted word or the compound word in the patent search database unit in association with the importance degree. A patent search device, comprising: a registration unit that registers as a term. 6. The patent search device according to claim 1, wherein the search unit is based on a distance between a vector corresponding to the search term phrase and a vector corresponding to the search condition phrase. A patent search device for searching patent documents having high relevance. 7. The patent search device according to claim 1, wherein the patent document includes descriptions of a plurality of “claims”, and the search unit includes a plurality of the search terms based on the search term phrase. A patent search device comprising a claim analysis unit for extracting a dependency relationship between "claims". 8. The patent search device according to claim 7, wherein the claim analysis unit includes a description of another “claim” included in the “claim” to be searched and the “claim” to be searched. And a patent search device for extracting the subordination relationship based on the title of the inventions in "Claims". 9. The patent search device according to claim 8, wherein the claim analysis unit includes the description of the other “claim” in the “claim” of the search target, and the “claim” of the search target. A patent search device, characterized in that when the title of the invention of claim is the same as the title of the invention of the other claim, it is determined that there is the subordinate relationship. 10. The patent search device according to claim 1, wherein the patent document includes a description of “claim”, and the search unit performs the search based on the search term phrase. A patent search device comprising a claim analysis unit for extracting a constituent feature included in the description of "claim". 11. A control method of a patent search apparatus having a patent search database section for storing a word or a compound word included in a patent document to be searched as a search term phrase in advance in association with the patent document. And a phrase extraction process of extracting a word or a compound word included in the condition setting patent document as a search condition phrase, comparing the search term phrase and the search condition phrase with respect to the search source patent document A method for controlling a patent search device, comprising: a search process for searching a patent document having a high similarity as a patent document highly related to the search source patent document. 12. The patent search device control method according to claim 11, wherein a reference relationship between the search source patent document and the highly related patent document is extracted based on the description in each patent document. A control method for a patent search device, comprising a reference destination extraction process. 13. The control method of the patent search device according to claim 12, wherein the patent search database unit relates to patent documents having a reference relationship in association with the search source patent document based on the extracted reference relationship. And a reference destination database unit for storing information on the reference source database unit, and based on the information in the reference destination database unit, creates a chart showing a reference relationship between the search source patent document and the highly related patent document. A method of controlling a patent search device, comprising: a process; and a display process for presenting the chart created by the chart creating unit. 14. The method of controlling a patent search device according to claim 11, further comprising: a morphological analysis process of performing morphological analysis of the search source patent document to extract the word or the compound word. An importance calculation process for calculating the importance of the extracted word or compound in the search source patent document, and the patent search database unit in which the extracted word or compound is associated with the importance. And a registration process for registering the search term as the search term. 15. The control method for a patent search device according to claim 11, wherein in the search step, a distance between a vector corresponding to the search term and a vector corresponding to the search condition term. A method for controlling a patent search device, which searches for patent documents having high relevance based on the above. 16. The method of controlling a patent search device according to claim 11, wherein the patent document includes a plurality of “claims”, and the search process is based on the search term phrase. A method for controlling a patent search device, comprising a claim analysis process for extracting a dependency relationship between the plurality of "claims". 17. The control method of the patent search device according to claim 16, wherein the claim analysis process includes the description of other “claims” included in the “claim” of the search target and the “claim” of the search target. A method for controlling a patent search device, characterized in that the subordinate relationships are extracted based on the titles of the inventions of "claim" and other "claims". 18. The control method of the patent search device according to claim 17, wherein in the claim analysis process, the “claim” of the search target includes the description of the other “claim”, and the search is performed. A control method of a patent search device, wherein when the name of the subject “claim” of the invention is the same as the name of the other “claim” of the invention, it is determined that there is the dependency. 19. The method for controlling a patent search device according to claim 11, wherein the patent document includes a description of “claim”, and the search process includes the search term in A method of controlling a patent search device, comprising a claim analysis process for extracting a constituent element included in the description of "claim" based on the claim. 20. A control program for causing a computer having a patent search database unit, which stores in advance a word or a compound word included in a patent document to be searched as a search term phrase in association with the patent document, to function as a patent search device. By analyzing the search source patent document, and extracting a word or a compound word included in the condition setting patent document as a search condition phrase, comparing the search term phrase and the search condition phrase, to the search source patent document On the other hand, a control program for searching a patent document having a high similarity as a patent document having a high relevance to the search source patent document. 21. The control program according to claim 20, wherein a reference relationship between the search source patent document and the highly related patent document is extracted based on the description in each patent document. And control program. 22. The control program according to claim 20 or 21, wherein the patent document includes a plurality of "claims", and the plurality of "claims" are mutually based on the search term phrase. A control program characterized by extracting subordinate relationships between them. 23. The control program according to claim 20 or 21, wherein the patent document includes a description of “claim” and is included in the description of “claim” based on the search term phrase. A control program characterized by extracting requirements. 24. A recording medium on which the control program according to any one of claims 20 to 23 is recorded.