JP2010122932A - Document retrieval device, document retrieval method, and document retrieval program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To add a proper score to a document by finding a proper related query to a query. <P>SOLUTION: A retrieval responding means 6 stores a log obtained by associating a query in a retrieval history of a user terminal 2 with a document accessed for a retrieval result in a click log database 8. A click log analyzing means 9 accumulates access degrees of an electronic document accessed for the retrieval result in a log, and stores the access degrees in an analysis result database 10. A query relevancy specifying means 11 uses an inclusion relation between queries included in the log to specify a related query, and determines a relevancy between the both queries to store the relevancy in a query relevancy database 12. A query document relevancy calculating means 13 uses relevancy between the both queries and an access degree from the a related query to calculate a score of the document, and stores the score in a document database 5. A retrieval executing means 7 responds to the user terminal 2 with a retrieval result reflecting the score. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a technique for searching an electronic document corresponding to a query from a group of electronic documents.

  In a system for searching a document collection such as a Web page, a document is scored based on a user's judgment independently of the contents of the document, and used for ranking search results.

  As the simplest method, there is a method of weighting useful documents manually. However, when the number of documents becomes enormous, it is very impractical to add completeness, and it is not practical.

Therefore, as a technique, the search engine log (search log) is analyzed, and the relationship between the query entered by the user and the Web page (document) of the search result that has been reached and the relationship between the queries are used in the past. Non-Patent Document 1 proposes a method for preferentially presenting web pages browsed in the past.
Ricardo Baeza-Yates et al. : "Query Recommendations Usage Query Logs in Search Engines", Proc. of ECIR 2005, pp. 588-596 Junya Murata, Hiroyuki Toda, Yumiko Matsuura, Ryoji Kataoka, “Proposal for Query Expansion Using Access Concentrated Sites in Search Results”. Symposium on Database and Web Information System (DBWeb2007): 2007.

  In the above-described method, the similarity between search results obtained from queries is used to evaluate the similarity between queries. However, in such an evaluation method, if the query is not sufficiently detailed, the topic included in the search result is also ambiguous, and a related query cannot always be found.

  Therefore, in view of such a problem, the present invention has a problem to find an appropriate related query for a query and to give an appropriate score to a document.

  In order to solve the above-described problem, the present invention specifies a related query and a related degree based on an inclusion relation and a replacement history of a query included in log information, and uses the related degree and an access degree from the related query as a document. The score is calculated.

  Specifically, the invention described in claim 1 is a document search device that searches for an electronic document including a query instructed by a user terminal, and includes a query included in a search history and a search result of the query. A storage unit that stores a log that is a combination of the accessed electronic documents, a totaling unit that totalizes the access degrees of the accessed electronic documents with respect to a query search result included in the log, and the log Relevance specifying means for identifying a related query using an inclusion relationship between the queries and a replacement history, and obtaining a relevance between the two queries, a degree of access from the related query, and a relevance between the two queries. And calculating means for calculating the score of the electronic document, and responding to the user terminal with a search result reflecting the score.

  The invention according to claim 2 is characterized in that the aggregation means obtains the usage frequency of the query included in the log, and the relevance specifying means obtains a ratio of the usage frequencies of the two queries as the association degree. It is characterized by.

  The invention according to claim 3 is characterized in that the relevance specifying means further obtains the degree of replacement between the two queries, and obtains the degree of association based on a ratio between the degree of substitution and the usage frequency. It is said.

  According to a fourth aspect of the present invention, there is provided a document search method for searching an electronic document including a query instructed from a user terminal, wherein the query included in the search history and the search result of the query are accessed. A first step of storing in the storage means a log combined with the electronic document that has been recorded, and a second step in which the counting means totalizes the access degree of the accessed electronic document with respect to the query search result included in the log And a third step in which the relevance specifying means specifies a related query using an inclusion relation and replacement history between the queries included in the log, and obtains a degree of relevance between the two queries, and a calculating means includes the And a fourth step of calculating the score of the electronic document using the degree of access from the related query and the degree of relevance between the two queries, and the search execution means reflects the score. It is characterized in that the response result with a user terminal.

  In the invention according to claim 5, the second step obtains the use frequency of the query included in the log, and the third step obtains the use frequency ratio of the two queries as the relevance. It is characterized by.

  The invention according to claim 6 is characterized in that the third step further obtains the degree of replacement between the two queries, and obtains the degree of association based on a ratio between the degree of substitution and the usage frequency. Yes.

  The invention described in claim 7 is a document search program, characterized in that a computer functions as each means constituting the document search device according to any one of claims 1 to 3.

  According to the first to seventh aspects of the present invention, an appropriate related query can be found by using the inclusion relationship between the queries included in the log information and the replacement history.

  By using the relevance between the two queries and the access from the related query, an appropriate score is given to the document, and such a score is reflected in the search result. In this respect, the search accuracy is improved.

  FIG. 1 shows a document search apparatus 1 according to an embodiment of the present invention. The document search apparatus 1 is connected to a user terminal (PC) 2 that instructs a search condition (query) via a network and a content server S in which an electronic document group is stored. Note that a plurality of user terminals 2 are usually connected to the document search apparatus 1.

  Here, it is assumed that the document search device 1 is configured as a server (for example, a search engine) that searches for content and the like existing in the content server S on the Internet. The document search apparatus 1 may be, for example, a computer that can be connected to a network and can execute processing logic for document search. The document search apparatus 1 is connected to a network other than the Internet such as an in-house LAN (Local Area Network). May be.

  The user terminal 2 includes a browser 3 as an interface with the user. The user terminal 2 may be a mobile terminal such as a mobile phone as long as the user terminal 2 can communicate with the document search apparatus 1 and includes the browser 3.

  As shown in FIG. 1, the document search apparatus 1 mainly functions as a search engine, and is a crawler 4, which is an information collection means 4, a document database 5, a search response means 6, a search execution means 7, a click log database 8, a click A log analysis unit 9, an analysis result database 10, a query relevance specifying unit 11, a query relevance database 12, and a query / document relevance calculation unit 13 are provided.

  The crawler 4 accesses the content server S through the communication device, collects electronic document groups to be searched, and stores the electronic document groups in the document database 5.

  The search response means 6 receives a query input by the user through the browser 3 from the user terminal 2 and transmits the query to the search execution means 7.

  The search execution means 7 searches the document database 5 with the query received from the search response means 6 and returns the acquired list of documents as a search result to the search response means 6. The search execution means 7 is realized by being incorporated in a search engine program or the like.

  The search response means 6 returns the search result received from the search execution means 7 to the user terminal 2. Then, a log obtained by combining the information of the document viewed by the user with respect to the search result and the query input at the time of the search is stored in the click log database 8 as a click log. This search response means 6 is implemented by being incorporated in a search application or the like.

  The click log analyzing means 9 analyzes the click log stored in the click log database 8, obtains the query included in the click log and the frequency of use thereof, and stores it in the analysis result database 10. At the same time, the document reached by the user in the search by each query and the degree of arrival are obtained and stored in the analysis result database 10.

  The query relevance specifying means 11 uses the storage information of the analysis result database 10 and the click log database 8 to obtain the degree of relevance between queries included in the click log and stores it in the query relevance database 12. .

  The query / document relevance calculation means 13 calculates the relevance (score) between a query and a document using the stored information of the query relevance database 12 and the analysis result database 10, and uses this to calculate the document database 5 To store.

  When the search instruction is received from the user terminal 2 after the score calculation, the search execution unit 7 ranks the documents according to the score, and uses the ranked list of documents as a search result via the search response unit 6. To the user terminal 2.

  The functions of the functional blocks 4 to 13 are realized by reading a document search program by a control unit (CPU: Central Processor Unit) of the document search apparatus 1. The document search device 1 includes normal components of a computer, for example, an input unit such as a keyboard and a mouse (not shown), a rewritable memory (RAM) for temporarily storing processing data, the user terminal 2 and A communication device used for network connection with the content server S, a storage unit such as a hard disk drive, and a display unit such as a display are provided. Of these, each of the databases 5.8.10.12 is constructed on the hard disk drive. The document database 5 may be implemented in the content server S. In this case, the connection to the document database 5 is made via the communication device. Hereinafter, the score calculation processing executed by each of the functional blocks 4 to 13 will be described with reference to the flowchart of FIG.

  S01: First, the crawler 4 collects electronic document groups to be searched from the content server S via a network and stores them in the document database 5.

  A data example of the document database 5 is shown in Table 1. In step S01, document information (document ID, title, text) is stored in the document database 5, and the score of each document is stored in step (S05) described later.

  Thereafter, the document search apparatus 1 accepts a search instruction from the user for the electronic document group stored in the document database 5. That is, when receiving the access from the user terminal 2, the search response means 6 presents a search interface screen to the browser 3, receives a query input by the user from the screen via the network, and receives the query. It transmits to the search execution means 7.

  The search execution means 7 searches the document database 5 with the received query, and returns a list of documents corresponding to the query to the search response means 6 as a search result.

  The search response means 6 returns the received search result to the user terminal 2 and presents it to the user via the browser 3. The user selects an arbitrary document from the presented search results and browses the document.

  S02: The search response means 6 stores in the click log database 8 a log that combines a user search history and an access history for the search result.

  That is, the search response means 6 associates the query input by the user at the time of search with the information (document ID, etc.) of the document actually viewed by the user for the search result based on the query, and clicks the click as a click log. Store in the log database 8. A data example of the click log database 8 is shown in Table 2. Note that the ID or the like of the document viewed by the user may be acquired from the browsing history of the user terminal 2.

  S03: The click log analyzing means 9 reads the click log from the click log database 8 and analyzes the following two points.

(1) Aggregation of Query Usage Frequency The click log analysis means 9 aggregates the usage frequency of each query included in the read click log, and stores the aggregation result in the analysis result database 10 as usage frequency information. Table 3 shows an example of data of usage frequency information in the analysis result database 10.

(2) Tabulation of arrival pages by query The click log analysis means 9 tabulates the documents (pages) actually reached by the user and the degree of their arrival in the search results for each query included in the click log. The result is stored in the analysis result database 10 as query arrival page information.

  The simplest example of the degree of achievement is a method using the number of accesses to a document. However, the simple access count includes a bias that the number of documents presented at the top of the search result is larger. Therefore, as in Non-Patent Document 2, a relative value such as the relationship with the previous and subsequent sites in the search result ranking may be used as an index representing the degree of achievement.

  Further, in order to exclude the access history to the document by the information collecting robot of the external search site from the click log database 8, a mechanical access pattern may be found and invalidated. Here, an example in which the number of accesses to a document is used as the degree of achievement will be described. Table 4 shows a data example of the arrival page information classified by query in the analysis result database 10 at this time.

  S04: The query relevance specifying means 11 specifies the relationship between the queries included in the usage frequency information of the analysis result database 10.

  Here, the relation between the queries in which the information request of the user who issued the query is similar is specified using the inclusion relation of the query. The inclusion relationship here refers to a case where there is another query that completely includes one query such as “Yokosuka” and “Yokosuka City Hall”. If the degree of association between these two queries is assumed to be degree of association 1, the degree of association 1 is calculated based on the following formula (1), for example. Note that the calculated relevance 1 data may be stored in the memory (RAM) or the storage unit.

  At this time, the replacement history of the query is used as an index for supplementing the degree of association 1 described above. The replacement history here refers to a replacement history between queries in which the user searches with the query “Yokosuka” and then searches again with the query “Yokosuka City Hall”.

  That is, the query relevance specifying unit 11 analyzes the click log of the click log database 8 and obtains the degree (replacement rate) at which the query is replaced as described above. At this time, when the user who inputs “Yokosuka” and the user who inputs “Yokosuka City Hall” are the same, it may be determined that the query has been replaced.

  If the degree of association between both queries based on the degree of substitution (replacement rate) is the degree of association 2, the degree of association 2 can be calculated by the following equation (2), for example. The calculated relevance level 2 data may be stored in the memory (RAM) or the storage unit.

The query relevance specifying means 11 calculates the relevance between the query and the related query using the relevance 1 and the relevance 2 determined in this way, and stores them in the query relevance database 12. This degree of association can be calculated by the following equation (3), for example. Here, q ₁ and q ₂ represent queries. Table 5 shows an example of data in the query relevance database 12 at this time.

  S05: The query / document relevance calculating means 13 calculates the relevance (score) between the query and the document using the stored information of the query relevance database 12 and the analysis result database 10.

That is, the query / document relevance calculation means 13 refers to the query relevance database 12 and reads the relevance of the query q _i and the related query q _x . Further, by referring to the analysis result database 10, reads the arrival degree to documents d _j from the search results by the relevant query q _x. Then, using these, for example, the query / document relevance (score) is calculated by the following equation (4), and the calculated score is stored in the document database 5 of Table 1. In addition, Q _rel _ _to _ _qi represents a set of related queries of query q _i.

  According to Table 1, the relevance (score) between the document whose document ID is “http://yokosuka.city.jp/” and the query “Yokosuka” is “1000”, and the relevance between the query “Yokosuka City Hall” ( The score) is “2000”. Expressions (1) to (4) are defined in the program of the document search apparatus 1.

  The score calculated in this way is used for document ranking when generating search results for subsequent search instructions. That is, when a search instruction from the user terminal 2 is received after the score is calculated, the search execution means 7 ranks the documents according to the degree of relevance (score) or similarity between the query input by the user and each document. . Then, the number of documents specified by the user is selected from the top of the ranking, and a list of the selected documents is returned to the search response means 6 as a search result. The search response means 6 returns the received search result to the user terminal 2 and presents it to the user via the browser 3.

  Note that the score calculation process described in S01 to 05 may be performed at regular intervals, whereby the latest score is reflected in the search result.

  As described above, according to the document search device 1 according to the present embodiment, since the inclusion relationship between the queries included in the click log and the replacement history are used, a query that is not sufficiently detailed is used. Can also find appropriate related queries.

  In addition, since the score of the document is obtained by using the degree of association between both queries and the degree of achievement from the related query, a high score is given to the document that the user who inputs the query may desire, It can be preferentially presented.

  In addition, this invention is not limited to the said embodiment, For example, it can provide also as a document search program which makes a computer function as each functional block 4-13 of the said document search device 1. FIG. This program may cause the computer to realize all the functions of the functional blocks 4 to 13 or may realize a part of the functions.

  This program is provided to the computer by downloading from a website or the like. The program is stored in a recording medium such as a CD-ROM, DVD-ROM, CD-R, CD-RW, DVD-R, DVD-RW, MO, HDD, Blu-ray Disk (registered trademark). It may be provided to a computer.

1 is a configuration diagram of a document search apparatus according to a first embodiment of the present invention. The processing flow figure of the same score calculation.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Document search device 2 ... User terminal 3 ... Browser 4 ... Crawler (information collection means)
5 ... Document database 6 ... Search response means (search application)
7 ... Search execution means (search engine)
8 ... Click log database (storage means)
9. Click log analysis means (counting means)
DESCRIPTION OF SYMBOLS 10 ... Analysis result database 11 ... Query relevance identification means 12 ... Query relevance database 13 ... Query and document relevance calculation means S ... Content server

Claims (7)

A document search device for searching for an electronic document including a query instructed from a user terminal,
Storage means for storing a log combining a query included in a search history and an electronic document accessed for the search result of the query;
A counting means for counting the access degree of the electronic document accessed for the search result of the query included in the log;
Relevance specifying means for specifying a related query using an inclusion relationship and replacement history between queries included in the log, and obtaining a degree of relevance between the two queries;
Using a degree of access from the related query and a degree of relevance between the two queries, and a calculating means for calculating a score of the electronic document,
A document search apparatus that responds to a user terminal with a search result reflecting the score. The aggregation means obtains the usage frequency of the query included in the log,
2. The document search apparatus according to claim 1, wherein the relevance specifying unit obtains a ratio of the usage frequencies of the two queries as the relevance. The relevance specifying means further obtains the degree of replacement between the two queries,
3. The document retrieval apparatus according to claim 2, wherein the degree of association is obtained based on a ratio between the degree of substitution and the frequency of use. A document retrieval method for retrieving an electronic document including a query instructed from a user terminal,
A first step of storing in a storage means a log combining a query included in the search history and an electronic document accessed for the search result of the query;
A second step in which the counting means counts the access degree of the electronic document accessed for the search result of the query included in the log;
A third step in which a relevance specifying unit specifies a related query using an inclusion relationship and a replacement history between the queries included in the log, and obtains a relevance between the two queries;
A calculating unit includes a fourth step of calculating a score of the electronic document using the degree of access from the related query and the degree of association between the two queries;
A document search method, wherein the search execution means responds to the user terminal with a search result reflecting the score. The second step obtains the usage frequency of the query included in the log,
5. The document search method according to claim 4, wherein the third step obtains a ratio of the usage frequencies of the two queries as the relevance. The third step further determines the degree of replacement between the two queries;
6. The document search method according to claim 5, wherein the degree of association is obtained based on a ratio between the degree of substitution and the usage frequency.   A document search program for causing a computer to function as each means constituting the document search device according to claim 1.

Images