JP4357820B2 - Concept search system and computer program - Google Patents

Description

[0001]
[Technical field to which the invention belongs]
  The present invention is a search system that applies a concept search using a vector operation to a search target document group stored in a certain database, and compensates for the drawbacks of a general concept search.
[0002]
[Prior art]
  (Concept search technology)
  The concept search technique is as follows. That is, a word is cut out from a dictionary creation document in advance by morphological analysis, and a basic vector is assigned to each word to generate a vector group (stem vector) having a dictionary function. Based on the stem vector, a vector of the entire search target document is determined, and a search target document vector group is generated from the search target document group. As a result, each search target document has one vector per document. When performing a search, a search text vector is generated based on the stem vector from the text input for the search, the inner product of the search text vector and the search target document vector group is calculated, and the inner product value is calculated. A predetermined number of documents are output from a high search target document.
  According to this technology, if a sentence specifying a target to be extracted (hereinafter referred to as “search sentence”) is input as a natural sentence, a predetermined number of documents are always output in order from documents with a high degree of relevance. It is possible to overcome the drawbacks of keyword search.
[0003]
[Patent Document 1]
    Japanese Patent Application No. 2000-197027
[0004]
  This patent document describes a search target document (specifically, technical information and patent information) that is a dynamic search target document group that is frequently updated or added and has a large number of words per document. A technique for achieving concept search is disclosed.
  (Similar search)
In the above-mentioned patent document, a search method defined as “similar search” that can omit the input of search text when performing concept search continuously using documents extracted as a result of concept search Is also disclosed.
[0005]
[Problems to be solved by the invention]
  When homonyms are included in the search sentence, a high score is given to a document that the searcher does not need.
  There is the term “big bang” as a homonym, but it was originally an astronomy term, and it was used figuratively in the financial industry as “financial big bang”. Today, it is often referred to simply as the “Big Bang”, and the proportion of people who associate astronomy terms is lower. When this “big bang” is included in the text for search, a high score is obtained for technical documents in the astronomical field and technical documents in the financial field.
[0006]
  The above problem cannot be solved even if a similar search is used. For example, it is assumed that a technical document in the field of financial engineering is extracted by performing a concept search by including “Big Bang” in the search text. This is because even if the technical document is specified and a similar search is performed, the term “Big Bang” included in the technical document is still an astronomy term.
[0007]
  In general, homonyms exist across multiple fields. That is, there are few terms that are used ambiguously in the same field, and in many cases, the terms have different meanings in different fields.
  On the other hand, when a searcher wants to perform a concept search or a similar search using a technical term, the searcher does not often recognize that the technical term is a homonym. In the above-described example, the concept search is performed without recognizing that “Big Bang” is an astronomy term, and the search result is noticed.
[0008]
  Now, if a system capable of performing concept search by field is provided, the features of concept search may be diluted. That is, it is because the feature of the concept search is that documents that are similar as a concept can be extracted even if crossing fields.
[0009]
  The problem to be solved by the present invention is to provide a concept search method in which a searcher can specify a field to be excluded from a search result.
  It is an object of the present invention to provide a concept search system that can specify a field to be excluded from search results.
  Another object of the present invention is to provide a concept search program that can specify a field to be excluded from search results.
[0010]
[Means for Solving the Problems]
  As means for solving the above problems, the present invention provides the following means.
  That is, the documents to be searched are classified in advance, a search target document vector group that is a basis for vector calculation is prepared for each field based on the classification, and the search target document vector group in a field other than the field to be excluded is used. This is a technique for deriving search results.
[0011]
  (Claim 1)
  According to the first aspect of the present invention, there is provided a stem vector creating means for creating a stem vector having a dictionary function from a dictionary creating document group, and a section or a section from a search target document group which is a collection of the stem vector and a document as patent information. Search target document vector generating means for generating a search target document vector group for each class, search data receiving means for receiving search data including search text input by a client, and the search text and stem vector A search sentence vector creating means for creating a search sentence vector by using, a vector calculation means for calculating a vector operation value of the search sentence vector and a search target document vector group, and a search in the order of the calculated vector operation value Sort means for sorting the target document group and outputting the sort result to the client. According to the example was the concept retrieval system.
  Each search target document of the search target document group includes data related to a section or class based on an international patent classification, and the search data is to be excluded from the search target document group.International patent classificationIncluding the exclusion field data which is a predetermined section or class of the international patent classification for identifying the search object document vector group other than the search object document vector group corresponding to the exclusion field data. This is a concept search system that extracts a vector group and calculates a vector operation value using the extracted search target document vector group and a search text vector.
[0012]
  (Glossary)
  The “client” is a terminal device related to a searcher who performs a search using the concept search system according to the present invention.
  The “field” refers to a group of classifications classified by a general classification method of a search target document group or a classification method based on an academic viewpoint. For example, when the document group to be searched is patent information, there are an international patent classification (so-called IPC) assigned to sections A to H, a class further classified into each section, and the like. In general, when a class is further reclassified, each class divided by the reclassification is called a subclass. However, in this specification, when the expression “class, subclass” is used, A classification indicates a class, and a subordinate classification indicates a subclass.
  “Classification data” includes not only specifying one of each field but also specifying a plurality of fields when there are n fields. In addition, there may be cases where a document to be classified extends over a plurality of classifications, and when the classification is permitted, it is possible to specify conditions over a plurality of fields. Specifically, in the IPC, when it is desired to exclude a document spanning the G section and the H section from the search target, the exclusion classification data “G and H” can be input.
  The “stem vector creation means” may create a stem vector for the entire dictionary creation document group, or create a stem vector for each of the divided fields in the dictionary creation document group. Good. When a stem vector is created for each field, a search target document vector group is created using the stem vector of that field, so that more appropriate search target document vectors and search text vectors can be created. it can.
  In addition, when an identification code or the like for specifying a field is given to the search target document in advance, when the search target document vector creation unit reads the search target document, the search target document The search target document vector may be created using an appropriate stem vector after judging the field.
[0013]
  (Function)
  The stem vector creating means creates a stem vector having a dictionary function. The search target document vector creation means includes a search field in the stem vector and the search target document group,Based on international patent classificationA search target document vector group is created for each of the divided fields.
  The search data receiving means receives search data input by the client. The search data includes search text and exclusion field data for specifying a field to be excluded from the search target. Using the search text and the stem vector, the search text vector creating means creates a search text vector. The vector calculation means calculates a vector calculation value using a search target document vector group and a search text vector based on field data other than the exclusion field data.
  The sorting means is a search target document group in the order of the calculated vector operation value.(Patent information)Sort and output the sort result to the client. As an output result, a search target document group extending over field data other than the excluded field data is output. In other words, from the fields that you want to exclude while taking advantage of the concept search, search target documents(Patent information)Is not extracted.
[0014]
  (Claim 2)
  The invention according to claim 2Patent information pre-classified into sections or classes based on international patent classificationThe present invention relates to a concept search system for executing a concept search on a search target document group.
  A stem vector creating means for creating a stem vector having a dictionary function from the dictionary creating document group, and a stem vector andEach section or class based on international patent classificationFrom the divided search target documents,Section or classA search target document vector creating means for creating a search target document vector group for each, a search data receiving means for receiving search data input by a client, and a vector for calculating a vector operation value using the search target document vector group Computation means and sorting means for sorting the search target document group in the order of the computed vector computation value and outputting the sort result to the client.
  I want to exclude search data from search target documents.A prescribed section or class of international patent classificationThe vector operation means extracts a search target document vector group other than the search target document vector group corresponding to the exclusion field data. Then, the vector calculation value is calculated using the extracted search target document vector group and the search text vector.
[0015]
  (Glossary)
  The invention described in claim 2 is different from the invention described in claim 1 in that it does not include “search sentence vector creating means”.
  Here, “literature data” refers to the document number or the like if there is a number or the like that centrally manages the search target document. For example, in the case of patent information, it is an application publication number.
  The “stem vector creation means” may create a stem vector for the entire dictionary creation document group, or create a stem vector for each of the divided fields in the dictionary creation document group. Good. When a stem vector is created for each field, a search target document vector group is created using the stem vector of the field, so that a more appropriate search target document vector can be created. In addition, when an identification code or the like for specifying a field is given to the search target document in advance, when the search target document vector creation unit reads the search target document, the search target document The search target document vector may be created using an appropriate stem vector after judging the field.
[0016]
  (Function)
  The stem vector creating means creates a stem vector having a dictionary function. The search target document vector creating means includes the stem vector and the search target document group.(Patent information)Search fields inInternational patent classification section or classA search target document vector group is created for each.
  The search data receiving means receives search data input by the client. A search sentence is specified using document data for specifying a search sentence in the search data. The vector calculation means calculates a vector calculation value using the specified search text and a search target document vector group based on field data other than the exclusion field data.
  The sorting means sorts the search target document group in the order of the calculated vector operation value, and outputs the sort result to the client. The output result is other than excluded field data.International patent classification section or classSearch target document group across(Patent information)Is output. In other words, from the fields that you want to exclude while taking advantage of the concept search, search target documents(Patent information)Is not extracted. If data for specifying the text for search is input, it is not necessary to input the text itself.
[0017]
  The following inventions can also be provided.
That is,A stem vector creating means for creating a stem vector having a dictionary function from the dictionary creating document group, and a stem vector andPatent information pre-classified for each section or class based on international patent classificationSearch target document vector generating means for generating a search target document vector group from the search target document group, search data receiving means for receiving search data including search text input by a client, search data and stem vector A search text vector creating means for creating a search text vector by using a vector, a vector calculation means for calculating a vector calculation value of the search text vector and a search target document vector group, and a vector calculation value of the calculated vector calculation value The present invention relates to a concept retrieval system comprising sorting means for sequentially sorting a search target document group and outputting a sorting result to a client.
  The search target document vector creation means is a search target document group.By section or class of international patent classificationA search target document vector group is created, and the sorting means includes a next search screen providing means for providing a next search input screen of the client, and the next search input screen is excluded for specifying a field to be excluded from the search target. Field dataFrom clientThe screen can be entered.The search data receiving means receives exclusion field data input from a client on the next search input screen.
  The vector calculation means includesIt is a predetermined section or class of the international patent classification specified by being input from the client to the next search input screen.A search target document vector group other than the search target document vector group corresponding to the exclusion field data is extracted from the search target document vector group, and the extracted search target document vector group andAboveRecalculate the vector operation value using the search text vector.
The sorting means outputs to the client a re-sorting result obtained by sorting the search target document group based on the international patent classification in the order of the re-calculated vector operation value. The concept retrieval system as described above.
[0018]
  (Glossary)
  After entering the search text only and obtaining the sort result, in the next search input screen, in addition to entering the exclusion field data, enter the search text and the class in the exclusion field data and enter the sort result. After obtaining the information, there is a case where a subclass is input as excluded field data on the next search input screen.
  The input on the next search input screen may be input by using a keyboard, selection by pull-down menu, or specification by clicking by a selectable button (hereinafter, including hyperlink).
  The “stem vector creation means” may create a stem vector for the entire dictionary creation document group, or create a stem vector for each of the divided fields in the dictionary creation document group. Good. When a stem vector is created for each field, a search target document vector group is created using the stem vector of that field, so that more appropriate search target document vectors and search text vectors can be created. it can. In addition, when an identification code or the like for specifying a field is given to the search target document in advance, when the search target document vector creation unit reads the search target document, the search target document The search target document vector may be created using an appropriate stem vector after judging the field.
[0019]
  (Function)
  The operation of inputting the exclusion field data that the client wants to exclude with respect to the first sort result of the concept search based on the search text, and outputting the second sort result again based on the exclusion field data will be described. . However, a further sort result can be output for the second sort result.
  A stem vector creating unit creates a stem vector having a dictionary function from the dictionary creating document group, and a search target document vector creating unit creates a search target document vector group from the stem vector and the search target document group. When the search data receiving means receives the search data including the search text input by the client, the search text vector creating means creates the search text vector by using the search data and the stem vector, and the search A vector calculation means calculates a vector calculation value of the text vector for search and the document vector group to be searched. Then, the search target document group is sorted in the order of the vector calculation values calculated by the sorting means, and the sort result is output to the client.
[0020]
  The sorting result includes the next search screen providing means for providing the next search input screen of the client, and the client uses the next search screen providing means to exclude the area to be excluded from the search target. Enter the field data.
  Using the input exclusion field data, the vector calculation means recalculates the vector calculation value using the search target document vector group and the search sentence vector based on the field data other than the exclusion field data. Then, the sorting means outputs the resort result obtained by sorting the search target document group based on the field data in the order of the recalculated vector operation value to the client again.
[0021]
  The following inventions can also be provided.
  That is, the invention includes a stem vector creation unit that creates a stem vector that performs a dictionary function from a dictionary creation document group, and a search target document vector creation unit that creates a search target document vector group from the stem vector and the search target document group. A search data receiving means for receiving search data input by the client, a search text vector creating means for creating a search text vector, and a vector operation value of the search text vector and the search target document vector group. The present invention relates to a concept search system comprising: vector calculation means for calculating; and sorting means for sorting search target document groups in the order of the calculated vector calculation values and outputting the sort result to a client.
  The search data includes emphasized word data that is a keyword that the client wants to emphasize, and includes search text editing means for editing the search text based on the emphasized word data, and the search text vector creating means includes: A search text vector is created using the search text and the stem vector edited by the search text editing means.
[0022]
  (Glossary)
  “Enhanced word data” refers to “Financial big bang” when, for example, “camera” is used as an emphasized word when a technology related to “film with lens” is extracted, or there is a homonym such as “big bang”. "Finance" is used as an emphasized word in order to extract documents closely related to "
  Further, ranks may be provided for the emphasized word data, or a magnification may be set. For example, automatic setting such as 3 times for a simple “emphasized word”, 5 times for a “superemphasized word”, and −1 time for a “negative word”, or a magnification can be appropriately input.
  The “search text editing means” is automatically performed by the server, not by the user who operates the client terminal. For example, the server performs operations such as rewriting a sentence that includes an emphasis word a plurality of times in the search sentence or repeatedly adding the emphasis word to the end of the search sentence.
[0023]
  (Function)
  A stem vector creation unit creates a stem vector that performs a dictionary function from the dictionary creation document group, and a search target document vector creation unit creates a search target document vector group from the stem vector and the search target document group.
  The search data receiving means receives the search data input by the client. The search data includes emphasized word data that is a keyword that the client wants to emphasize, and the search sentence editing means edits the search sentence based on the emphasized word data. The search text vector creating means creates a search text vector using the search text and the stem vector edited by the search text editing means, and a vector of the search text vector and the search target document vector group. The vector calculation means calculates the calculated value. Then, the search target document group is sorted in the order of the vector calculation values calculated by the sorting means, and the sort result is output to the client.
[0024]
  In the search result, a keyword to be emphasized is emphasized, and a search target document group extending over a plurality of field data is output. That is, it is possible to obtain a search result based on a search sentence in which a keyword to be emphasized is emphasized while taking advantage of the concept search.
[0025]
  The following inventions can also be provided.
  That is, the invention includes a stem vector creation unit that creates a stem vector that performs a dictionary function from a dictionary creation document group, and a search target document vector creation unit that creates a search target document vector group from the stem vector and the search target document group. Search data receiving means for receiving search data including search text input by the client, search text vector creating means for creating a search text vector using the search text and stem vector, and A vector calculation means for calculating a vector operation value of a search text vector and a search target document vector group, a sorting means for sorting the search target document group in the order of the calculated vector calculation value, and outputting a sort result to a client; Relates to a concept search system comprising:
  The sorting means includes a next search screen providing means for providing a next search input screen of the client. The next search input screen is a screen on which emphasized word data that is a keyword to be emphasized by the client can be input. A search text editing means for editing the search text based on the above. The search text vector creating means creates a re-search text vector using the search text and the stem vector edited by the search text editing means, and the vector calculation means is the re-search text vector and the search target The vector operation value with the document vector group is recalculated, and the sorting unit outputs a resort result obtained by sorting the search target document group based on the field data in the order of the recalculated vector operation value to the client.
[0026]
  (Function)
  A stem vector creation unit creates a stem vector that performs a dictionary function from the dictionary creation document group, and a search target document vector creation unit creates a search target document vector group from the stem vector and the search target document group. The search data receiving means receives the search data including the search text input by the client, and the search text vector creating means creates the search text vector using the search text and the stem vector. Then, the vector calculation means calculates the vector operation value of the search text vector and the search target document vector group, the sort means sorts the search target document group in the order of the calculated vector operation value, and the sorting result to the client Is output.
[0027]
  The sorting means includes a next search screen providing means for providing a next search input screen of the client, and the next search input screen is a screen on which emphasized word data that is a keyword to be emphasized by the client can be input. The search text editing means edits the search text based on the emphasized word data. The search text vector creating means creates a search text vector by using the search text and the stem vector edited by the search text editing means, and the vector calculation means has the re-search text vector and the search target. Recalculate the vector operation value with the document vector group. Then, the sorting unit outputs a resort result obtained by sorting the search target document group based on the field data in the order of the recalculated vector operation value to the client.
  Here, the first sorting result is examined, and the emphasized word data is input, and the vector operation is performed on the search sentence edited by the emphasized word data. As a result, it is possible to obtain a search result using a search sentence in which a keyword to be emphasized is emphasized while utilizing the merit of the concept search.
[0028]
(Delete)
[0029]
(Delete)
[0030]
(Delete)
[0031]
(Delete)
[0032]
  A method invention corresponding to claim 1, that is, a stem vector creation procedure for creating a stem vector having a dictionary function from a dictionary creation document group, and a search target document group obtained by dividing the stem vector and the search field into a plurality of parts Search target document vector creation procedure for creating a search target document vector group for each field, and search data including search text input by the client and exclusion field data for specifying a field to be excluded from the search target. A search data receiving procedure to receive, a search text vector creation procedure to create a search text vector using the search text and stem vector, and a search target document vector group based on field data other than the exclusion field data; A vector calculation procedure for calculating a vector calculation value using a search sentence vector, and a calculation It is also possible to provide a concept search method including a sort procedure for sorting a search target document group in the order of vector operation values and outputting a sort result to a client.
[0033]
  Further, a method invention corresponding to claim 2, that is, a stem vector creation procedure for creating a stem vector having a dictionary function from a dictionary creation document group, and a search target document group obtained by dividing the stem vector and the search field into a plurality of parts Search target document vector creation procedure for creating a search target document vector group for each field, and exclusion field data for specifying a field to be excluded from the search target input by the client and literature data for specifying the search text A vector calculation value is calculated based on a search data reception procedure for receiving search data including a search target document vector group based on field data other than exclusion field data and a search sentence specified by literature data. Sort the search target document group in the order of the vector calculation procedure and the calculated vector calculation value, It is also possible to provide a concept search method and a sorting procedure for outputting the sorted results to clients.
[0034]
  (Claim 3)
  Claim 3The invention described in FIG.Patent information pre-classified into sections or classes based on international patent classificationThe present invention relates to a program for executing a concept search on a search target document group.
  The program includes a stem vector creation procedure for creating a stem vector having a dictionary function from a dictionary creation document group, the stem vector andEach section or class based on international patent classificationFrom the search target document group divided intoSection or classA search target document vector creation procedure for creating a search target document vector group for each, a search data reception procedure for receiving search data including search text input by a client, and the search data and stem vector. Search text vector creation procedure for creating a search text vector, a vector calculation procedure for calculating a vector calculation value of the search text vector and the search target document vector group, and a search target in the order of the calculated vector calculation value Sorting a group of documents and outputting the sorting result to the clientA prescribed section or class of international patent classificationExclusion field data receiving procedure for receiving exclusion field data for identifying the search field document vector group other than the search target document vector group corresponding to the exclusion field data is extracted from the search target document vector group and extracted Recalculation procedure for recalculating the vector operation value using the search target document vector group and the search text vector, and the recalculated vector operation value in this order.International patent classificationA computer-readable program for causing a computer to execute a re-sorting procedure for outputting a re-sorting result obtained by sorting a search target document group based on the above to a client.
[0035]
(Delete)
[0036]
(Delete)
[0037]
  Claim 3It is also possible to provide the program described in 1 above by storing it in a recording medium. Here, the “recording medium” is a medium that can carry a program that cannot occupy space by itself, such as a flexible disk, a hard disk, a CD-ROM, an MO (magneto-optical disk), a DVD- ROM or the like.
[0038]
  It is also possible to provide a computer program invention corresponding to claim 1.
  The program is a computer-readable program for executing a concept search on a search target document group previously classified in each field by a general classification method or a classification method based on an academic viewpoint. ,
  A stem vector creation procedure for creating a stem vector having a dictionary function from a dictionary creation document group, and a search target document for creating a search target document vector group for each field from the search target document group divided into a plurality of search fields. A search data receiving procedure for receiving search data including a vector creation procedure, search text input by a client, and exclusion field data for specifying a field to be excluded from a search target document group, and the exclusion field data A search text vector creation procedure for creating a search target document vector group after removing the search target document group corresponding to the exclusion field data from the search target document group, and the search target document vector group and the search target A vector calculation procedure in which the vector calculation means calculates a vector calculation value using a sentence vector; This is a computer-readable program for causing a computer to execute a sorting procedure for sorting a search target document group in the order of the calculated vector operation values and outputting a sorting result to a client.
[0039]
  It is also possible to provide a computer program invention corresponding to claim 2.
  The program is a computer-readable program for executing a concept search on a search target document group previously classified in each field by a general classification method or a classification method based on an academic viewpoint. ,
  A stem vector creation procedure for creating a stem vector that performs a dictionary function from a dictionary creation document group, and a search target document vector group for each field from the search target document group obtained by dividing the stem vector and the search field into a plurality of fields. Search for receiving search data including search target document vector creation procedure, exclusion field data for specifying fields to be excluded from the search target document group input by the client, and literature data for specifying search text Data reception procedure, a vector calculation procedure for calculating a vector calculation value based on a search target document vector group based on field data other than exclusion field data and a search sentence specified by literature data, and a calculated vector calculation Sort the search target documents in the order of the values and output the sort result to the client A computer readable program for executing the steps on a computer.
[0040]
DETAILED DESCRIPTION OF THE INVENTION
  Hereinafter, the present invention will be described in more detail based on embodiments and drawings. The drawings used here are FIGS. 1 to 10. 1 to 5 are conceptual diagrams showing an embodiment. FIG. 6 is a conceptual diagram showing the principle of search text editing means used in the fourth and fifth embodiments. 7 to 10 are diagrams showing specific search result screens.
[0041]
  (Figure 1)
  FIG. 1 is a conceptual diagram showing a first embodiment of the present invention, which is a device provided in a server in a client / server system. A client user receives a member ID and a password based on a contract with a business entity that operates the server, connects to the Internet, and uses a service provided by the server.
[0042]
  The server includes a stem vector creation means for creating a stem vector having a dictionary function from the dictionary creation document group, and a search target document vector group for each field from the search target document group obtained by dividing the stem vector and the search field into a plurality of fields. Search target document vector creating means for creating search data receiving means for receiving search data input by the client, search sentence vector creating means for creating a search sentence vector using the search data, Vector calculation means for calculating a vector operation value of the search sentence vector and the search target document vector group, and a sort means for sorting the search target document group in the order of the calculated vector operation value and outputting the sort result to the client And.
[0043]
  This server can search for patent information, and both the search target document group and the dictionary creation document group are patent information.
  The stem vector creation means creates a stem vector in advance from a dictionary creation document group. Further, the search target document vector creating means has field information so that the search fields in the search target document group can be classified into three fields A, B, and C, and fields divided based on the field information. A search target document vector group is created in advance for each.
  In the present embodiment, since the search target document group is patent information, the international patent classification, and the class finely classified for each section correspond to the “field”. Here, for the sake of easy understanding, the field in the dictionary creation document group and the search field in the search target document group are described as being divided into three fields. However, more categories may be used, or a certain category. It is good also as a classification | category hierarchized, such as making the further classification exist in.
[0044]
  The search data input by the client through the search data input means includes search text and field data for a field to be excluded for specifying the search field. Specifically, the fact that “A” field is specified is included as a certain search sentence and field data.
  The search data receiving means receives a search text and exclusion field data (A). The search text vector creating means creates a search text vector using the stem vector of the field of the search text, and from the search target document vector group based on the received exclusion field data (A), Specifies a search target document vector group. Then, the vector calculation means calculates a vector calculation value by using the search target document vector group other than the A field and the search text vector.
[0045]
  The calculated vector calculation value is based on a search target document vector group other than the A field, and the search target document vector group other than the A field is based on the stem vector and the search target document group of the B and C fields. Has been created. For this reason, even if a word used in the search sentence includes a homonymous word representing another meaning in the A field, a search target document in the A field is not extracted. That is, it is possible to obtain a search result that takes advantage of the concept search while removing the search target to be excluded.
[0046]
  (Figure 2)
  FIG. 2 is a conceptual diagram showing a second embodiment of the present invention.
  The server shown here includes a stem vector creation means for creating a stem vector having a dictionary function from a dictionary creation document group, and a search target for each field from the search target document group divided into a plurality of stem vectors and search fields. A search target document vector creating means for creating a document vector group, a search data receiving means for receiving search data input by a client, a vector calculation means for calculating a vector operation value using the search target document vector group, Sorting means for sorting the search target document group in the order of the calculated vector operation value and outputting the sort result to the client is provided.
[0047]
  A significant difference from the first embodiment is that the search sentence vector creating means is omitted. The search data includes exclusion field data for specifying the search field and document data for specifying the search text. Based on the search text specified by the document data, one search target document vector specified from the search target document vector group created in advance by the search target document vector generation means is used as the search document vector. . The vector calculation means calculates a vector calculation value using a search target document vector group based on a field other than the field specified by the exclusion field data and a search document vector that is one specified search target document vector. We are going to calculate. The search method shown here is a kind of similar search that extracts a document having a concept similar to a search sentence specified by document data.
[0048]
  It is assumed that the client inputs document data and exclusion field data that are search data. Here, it is assumed that a document number (2000-123) such as a publication number capable of specifying a document is input as the document data, and a symbol (C field) for specifying the field is input as the excluded field data. The search data is received by the search data receiving means in the server. Then, a vector matching the document number (2000-123) is extracted from the search target document vector group.
  The vector calculation means performs vector calculation on the vector matching the extracted document number (2000-123) and the search target document vector group in the field other than the C field. The concept search can be performed without the presence or function of the search text vector creation means. As a result, the client can perform the concept search without inputting any search text, while the server side can reduce the burden on the hardware and can shorten the calculation time.
[0049]
  (Figure 3)
  FIG. 3 is a conceptual diagram showing a third embodiment of the present invention. In the system of this embodiment, a search is first performed using a search sentence, and a re-search is performed using excluded field data for specifying a field based on the search result.
  The server shown here includes a stem vector creation means for creating a stem vector having a dictionary function from a dictionary creation document group, and a search target for each field from the search target document group divided into a plurality of stem vectors and search fields. A search target document vector creating means for creating a document vector group, a search data receiving means for receiving search data including a search text input by a client, and a search text vector using the search data Search text vector creation means, vector calculation means for calculating a vector operation value of the search text vector and search target document vector group, and sort the search target document group in the order of the calculated vector operation value, Sorting means for outputting the sort result.
[0050]
  The stem vector creation means creates a stem vector from the dictionary creation document group. Further, the search target document vector creating means has field information so that the search fields in the search target document group can be classified into three fields A, B, and C, and fields divided based on the field information. A search target document vector group is created in advance for each. In the initial search, only the search text is input, and the field is not designated by the field data. Therefore, the vector calculation means performs vector calculation on the search target document vector group using the entire vector group and the search text vector. As a premise of this vector operation, an entire vector group that is not divided into fields may be created using stem vectors for all fields.
  The sorting means includes a next search screen providing means for providing a next search input screen of the client, although not shown in the drawing, and the next search input screen displays excluded field data for specifying a search field to be excluded. The screen can be entered using a pull-down menu.
[0051]
  The above-described vector calculation means recalculates the vector calculation value using the search target document vector group and the search sentence vector other than the exclusion field based on the exclusion field data input from the next search input screen. Similarly, the above-described sorting means also outputs to the client a re-sorting result obtained by sorting the search target document group based on the field data in the order of the re-calculated vector operation value.
[0052]
  Assume that the client provided with the next search input screen inputs and transmits the exclusion field data (A). A search target document vector group in a field other than A is extracted from the excluded field data (A). Then, the vector calculation means recalculates the vector calculation value. In this case, if a word used in the search sentence includes a homonym that represents another meaning in the A field, the document in the A field gives a high score in the first sort result. there is a possibility. However, since the resort result does not include the document in the A field, the search target document in the A field is not extracted. In other words, this embodiment is effective when, for example, it is noticed from the first sort result that a homonym with another meaning in the A field is included.
[0053]
  (Fig. 4)
  FIG. 4 is a conceptual diagram showing the fourth embodiment of the present invention, and shows a concept search method for searching by inputting emphasized word data (α) together with search sentences. In other words, in addition to the search text as search data, the keyword “emphasized word data (α)” that is to be emphasized at the time of search extraction is input. To get a good search output.
The search data receiving means receives both the search text and the emphasized word data (α). Then, the search text editing means taking into account the emphasized word data (α) is edited by the search text editing means. This search sentence editing is not an operation performed by the user operating the client terminal, but the server automatically performs an editing operation such as repeatedly emphasizing the emphasized word data (α) into the sentence. The search text vector creating means creates a search text vector based on the edited search text and a separately created stem vector. Then, the vector calculation means outputs a vector calculation value using the search text vector and the search target document vector group, and the sorting means outputs the sort result to the client.
[0054]
  (Fig. 6)
  The principle of the search text input means and the emphasized word input means constituting the search data input means will be described with reference to FIG.
  On the search data input screen, there are a column for entering “search text”, a column for entering “emphasized word”, a column for inputting “super-emphasized word” to be emphasized, and an opposite word or a word to be excluded. A column for inputting “negative word” is provided.
  In the “Search text” field, enter “Used car online auction.” In the “Highlight word” field, enter “Internet.” In the “Super-highlight word” field, enter “Automobile”. It is assumed that “new car” is entered in the “negative word” field.
[0055]
  In the lower left of FIG. 6, vectors formed by words such as “used car”, “auction”, “Internet”, and the like are shown. As shown in the lower right part of FIG. 6, the search sentence editing means performs editing such that the vector related to the emphasized word is tripled, the vector related to the super-enhanced word is multiplied by 5, and the vector related to the negative word is minus 1 . In the search sentence vector created after editing in this way, the emphasized word and the super emphasized word are emphasized three times and five times, and the negative word is reflected as a vector in the opposite direction.
[0056]
  (Fig. 5)
  FIG. 5 is a conceptual diagram showing the fifth embodiment of the present invention. First, a procedure is shown in which a concept search is performed only by using a search sentence, the emphasized word data (α) is input using the output result, and a search result utilizing the emphasized word data (α) is obtained.
  As for the first search text, the search text vector creating means creates the search text vector as it is, and a vector operation is performed using the search text vector and the search target document vector group to obtain a sorted result. When the client inputs the emphasized word data (α) on the “next search screen” obtained together with the sort result, the emphasized word data (α) is used by the search sentence editing means to edit the search sentence. A re-search text vector is created using the edited search text, and vector calculation is performed again using the search text. The vector operation value is re-sorted by the sorting means and output as a re-sort result.
[0057]
  The search shown in FIG. 5 experiences that the search result is blurred unless it is clearly stated that “net” is “Internet” in a search sentence such as “auction a used car online”. You can get search results that take advantage of that experience.
[0058]
  (Variation of combination)
  The concept search utilizing “exclusion field data” is described in FIGS. 1 to 3 and the concept search utilizing “emphasized word data” is illustrated independently in FIGS. 4 to 6. ”And“ emphasized word data ”can naturally be provided as a search data.
[0059]
  (Existence by field)
  In the embodiment shown in FIGS. 4 and 5, the stem vector and the search target document group are described as being created without dividing the field, but the stem vector and the search target document group may be created for each field. The spirit of the present invention can be realized.
[0060]
  (Fig. 7)
  FIG. 7 shows an output result of a simple concept search using a search sentence “Auction a used car online”.
  The "net" in the search text above is only conscious of "Internet" as a client, but also means the meaning of a physical "net", so "~ protection net and fishing net", "vehicle net" The patent application related to the invention name such as "" has also been output as a sort result.
[0061]
  (Fig. 8)
  FIG. 8 is a diagram showing a search result screen in which “Internet” is input as an emphasized word. According to this search result, the sort result related to the invention name which is expected to mean the meaning of physical “network” is not output. On the other hand, the percentage of applications that clients would have wanted to hit, such as “used car auction management device”, “auction mall system”, “used car buying and selling system”, definitely increased compared to the case of FIG. ing.
[0062]
  (Fig. 9)
  It is a figure which shows the screen of the search result which input "new car" as a negative word. Comparing this sort result with the case of FIG. 8, it is clear that the auction system for products not related to “used cars” has been removed and the ratio of applications with “used cars” in the name of the invention has increased. is doing.
[0063]
  (Fig. 10)
  FIG. 10 is a diagram showing a search result screen in which the search text “Auction a used car” and the international patent classification A classification as exclusion field data are input. It is considered that the sorting result targeted by the client could be obtained from FIG. 7 by inputting the A classification that seems to have many inventions meaning the physical “network” as exclusion field data.
[0064]
  In the embodiment described above, a search system in which the dictionary creation document group and the search target document group are the patent information document group is used. However, the present invention is not limited to the patent information search system.
  On the other hand, even in a patent information search system, it is possible to create a dictionary creation document group from, for example, newspaper information other than the patent information document group.
[0065]
【The invention's effect】
  According to the first to third aspects of the invention, it is possible to provide a concept search system in which a searcher can specify a field to be excluded from a search result.
  Further, according to the invention described in claim 4, it is possible to provide a concept search program that allows a searcher to specify a field to be excluded from a search result.
[Brief description of the drawings]
FIG. 1 is a conceptual diagram showing a first embodiment.
FIG. 2 is a conceptual diagram showing a second embodiment.
FIG. 3 is a conceptual diagram showing a third embodiment.
FIG. 4 is a conceptual diagram showing a fourth embodiment.
FIG. 5 is a conceptual diagram showing a fifth embodiment.
FIG. 6 is a conceptual diagram showing the principle of search text editing means used in the fourth and fifth embodiments.
FIG. 7 is a diagram showing a specific screen (first search screen) provided first.
FIG. 8 is a diagram showing a search result screen in which an emphasized word is input.
FIG. 9 is a diagram illustrating a search result screen in which a negative word is input.
FIG. 10 is a diagram showing a search result screen in which excluded field data for excluding a certain field is input.

Claims (3)

Stem vector creation means for creating a stem vector having a dictionary function from the dictionary creation document group, and a search target document vector group for each section or class from the search target document group that is a collection of the stem vector and the document that is patent information Search target document vector creation means for creating search data, search data reception means for receiving search data including search text input by the client, and creation of a search text vector using the search text and stem vector A search sentence vector creating means, a vector calculation means for calculating a vector calculation value of the search sentence vector and the search target document vector group, and sorting the search target document group in the order of the calculated vector calculation value, A concept retrieval system comprising sorting means for outputting a sorting result to a client. Te,
Each search target document of the search target document group includes data on a section or class based on an international patent classification,
The search data includes exclusion field data as a predetermined section or class of an international patent classification for specifying an international patent classification to be excluded from a search target document group,
The vector calculation means extracts a search target document vector group other than the search target document vector group corresponding to the exclusion field data from the search target document vector group, and the extracted search target document vector group, search text vector, A concept search system that uses vector to calculate vector operation values. A concept search system for executing a concept search on a search target document group that is a collection of documents that are patent information including data on sections or classes based on an international patent classification ,
A stem vector creating means for creating a stem vector forming a dictionary function from the dictionary creation documents, retrieved from stem vector and target document group that is divided into sections or classes based on the international patent classification per the section or class Search target document vector generating means for generating a target document vector group, search data receiving means for receiving search data input by a client, vector calculation means for calculating a vector operation value using the search target document vector group, Sorting means for sorting the search target document group in the order of the calculated vector operation value and outputting the sort result to the client,
The search data includes excluded field data for specifying a predetermined section or class of the international patent classification to be excluded from the search target document group, and literature data for specifying the search text,
The vector calculation means extracts a search target document vector group other than the search target document vector group corresponding to the exclusion field data, and uses the extracted search target document vector group and the search sentence vector to obtain a vector calculation value. A concept search system that is supposed to be operated. A program for causing a computer to execute a concept search on a search target document group that is a collection of documents that are patent information including data on sections or classes based on international patent classifications ,
The program includes a stem vector creation procedure for creating a stem vector having a dictionary function from a dictionary creation document group,
A search target document vector creation procedure for creating a search target document vector group for each section or class from the search target document group divided into each section or class based on the stem vector and the international patent classification ;
Retrieval data reception procedure for receiving search data including search sentences entered by the client;
A search text vector creation procedure for creating a search text vector using the search data and stem vector;
A vector operation procedure for calculating a vector operation value between the search sentence vector and the search target document vector group,
A sorting procedure for sorting the search target document group in the order of the calculated vector operation value and outputting the sort result to the client,
Exclusion field data reception procedure for receiving exclusion field data for identifying a predetermined section or class of an international patent classification to be excluded from search,
A search target document vector group other than the search target document vector group corresponding to the exclusion field data is extracted from the search target document vector group, and a vector operation value is calculated using the extracted search target document vector group and search sentence vector. Recalculation procedure to recalculate
A computer-readable program for causing a computer to execute a re-sorting procedure for outputting to a client a re-sorting result obtained by sorting a search target document group based on an international patent classification in the order of re-calculated vector operation values.

Images