JP5944867B2 - Search intention dictionary creation device, search intention dictionary creation method, and search intention dictionary creation program - Google Patents

Description

  The present invention relates to a search technique, and more particularly to a technique for classifying search keyword intentions.

  The simplest method for acquiring a search intention using a search log of a search engine is to first query a log database based on a keyword given by a user, and execute a query of a plurality of keywords including the keyword. I'm pulling out. And this query is counted up frequently, and what has a large number is specified as a search intention related to a keyword.

  Furthermore, there is a method of specifying a search intention related to a keyword based on information of a document selected from a document group acquired by the search when the search keyword is input. In this method, a record related to a specific search keyword is extracted from a query log, URLs obtained there are clustered, and a cluster obtained thereby is intended to be searched for the query (Non-Patent Document 1).

Yunhua Hu, “Mining query subtopics from search log data”, Proc.SIGIR'12, August 12-16, 2012, Portland, Oregon, USA.pp.305-314

  In the former method, it is possible to acquire the search intention related to the keyword only by counting the simple appearance frequencies. However, in this method, there is a problem that a plurality of items with high relevance between the obtained search intentions are extracted, and there is a problem when using this output result.

  On the other hand, the method of Non-Patent Document 1 can suppress extraction of similar search intentions by performing clustering using the URL of the click log.

  However, no matter which method is used, it is impossible to grasp whether the search intention obtained there is general or specific to the keyword.

  The present invention has been made in view of the above circumstances, and an object thereof is to provide a search intention dictionary that can more accurately obtain a user's search intention.

  Therefore, the present invention acquires search intention candidates based on the keywords provided for the search by the user, and uses these candidates to acquire keyword synonyms, and further extracts the search intention candidates of these synonyms. The search intention candidate as a class including the keyword is acquired. As described above, by performing analysis based on the keyword and the search intention candidate, it is possible to grasp whether the search intention is specialized for the keyword or the search intention shared by the whole class.

  An aspect of the present invention is an apparatus for creating a search intention dictionary, in which a search engine searches using a search query including a keyword provided for search and one or more keywords other than the keyword as a search query related to the keyword. Search query extraction means for extracting from the log, keyword search intention candidate specifying means for specifying a word from which the keyword has been deleted from the extracted search query as a keyword search intention candidate, and search for each of the specified search intention candidates A search query related to the candidate from the log, a word obtained by deleting the candidate part of the search intention from the search query, and a synonym acquisition means for acquiring a word selected from the set of words as a synonym; The obtained synonym is used for the search query extraction means to obtain a search query including the isotope. Acquires a word from which the synonym is deleted as a candidate for a search intention of the same word, and specifies a set of search intentions selected from all of the acquired candidates for a search intention as a search intention candidate related to the set of peer terms Search that classifies into general search intentions and search candidates specialized for the keywords based on the class search intention candidate specifying means to perform, and the search intention candidates for the keywords and the search intention candidates related to the set of the same terms Intention classification means.

  An aspect as a method of the present invention is a search intention dictionary creation method executed by a search intention dictionary creation device, wherein the search query extraction means of the device is used for a search and one or more keywords other than this keyword A search query including a keyword as a search query related to a keyword from a search log of a search engine, and a keyword search intention candidate specifying unit of the device deletes the keyword from the extracted search query and searches for a keyword And a synonym acquisition unit of the device acquires a search query related to the candidate from the search log for each of the specified search intention candidates, and the search intention candidate portion is determined from the search query. Totalizing deleted words, obtaining a word selected from the set of words as a synonym; The class search intention candidate specifying means provides the obtained query terms to the search query extraction means to obtain a search query including the synonyms, and a word obtained by deleting the synonyms from the search query. A step of identifying a set of search intentions acquired as candidates, and a set of search intentions selected from the acquired search intention candidates of all the same terms, and a search intention classification means of the device, And classifying the keyword into a general search intention and a search candidate specialized for the keyword based on the keyword search intention candidate and the search intention candidate related to the set of synonyms.

  Note that the present invention may be in the form of a program that causes a computer to function as each unit of the apparatus or a program that causes a computer to execute the steps of the method.

  ADVANTAGE OF THE INVENTION According to this invention, the search intention dictionary which can obtain a user's search intention more correctly can be provided.

The block block diagram of the search intention dictionary creation apparatus which is embodiment of this invention. An example of the input-output data in each function part of the embodiment. An example of the search log database of the embodiment. An example of the search intention database of the embodiment. The flowchart figure explaining the processing procedure of the embodiment.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. However, the present invention is not limited to these embodiments.

[Overview]
The search intention dictionary creation device 1 shown in FIG. 1 analyzes the search history of a search engine and specifies not only the search intention, but also whether the search intention is a keyword specialized only for the keyword or the like. Identify whether it is strongly related to the keywords of the class. According to this, for example, a highly specialized search intention can be used for recommendation or linked to special content. Moreover, if it can be grasped that it is general, it is possible to prepare a specialized database or the like that accepts such an intention search.

[Device configuration]
As shown in FIG. 1, the search intention dictionary creation device 1 includes a log filter unit 11, a keyword search intention candidate specifying unit 12, a synonym acquisition unit 13, a class search intention candidate specifying unit 14, and a search intention classification unit 15. Prepare.

  Each of the functional units 11 to 15 is realized by computer hardware resources. That is, the apparatus 1 includes hardware resources related to a computer such as at least an arithmetic device (CPU), a storage device (memory, a hard disk device, etc.), a communication interface, and the like. These hardware resources cooperate with software resources (OS, applications, etc.) to implement the functional units 11 to 15. Moreover, you may make it each implement | achieve the function parts 11-15 in each computer.

  The log filter unit 11 extracts a search query including a keyword used for a search and one or more keywords other than the keyword from a search log database 21 (hereinafter referred to as a search log DB 21) of a search engine as a search query related to the keyword. To do.

  The search log DB 21 is a database in which logs when a user uses a search engine are recorded. In addition to the time stamp, the search query entered by the user is recorded. FIG. 3 shows an example of the search log DB 21.

  The keyword search intention candidate specifying unit 12 specifies a word obtained by deleting the keyword from the extracted search query as a search intention candidate for the keyword.

  The synonym acquisition unit 13 acquires a search query related to the candidate from the search log DB 21 for each of the specified search intention candidates, and totals the words obtained by deleting the search intention candidate portion from each search query. The word selected from the set of is acquired as an equivalent word.

  The class search intention candidate specifying unit 14 supplies the acquired homologous word to the log filter unit 11 to acquire a search query including the homologous word, and selects a word from which the synonym is deleted from the search query as a candidate for a search intention of the homologous word And a set of search intentions selected from the acquired search intention candidates of all the same terms is specified as a search intention candidate related to the set of the same terms.

  The search intention classification unit 15 classifies a general search intention and a search candidate specialized for the keyword based on the search intention candidate of the keyword and the search intention candidate related to the set of the same terms. The classified search intentions are stored in a search intention database 22 (hereinafter, search intention DB 22).

  The search intention DB 22 is a database that stores the “general search intention” and the “characteristic search intention” specified by the search intention classification unit 15 in association with keywords and can be updated. An example of the search intention DB 22 is shown in FIG.

[Description of processing procedure]
The processing procedures S1 to S9 of the present embodiment shown in FIG. 5 will be described with reference to the cases of FIGS.

  S1: The log filter unit 11 receives an input of a keyword, accesses the search log DB 21 accordingly, and acquires a search log record including the keyword and a character string other than the keyword. The acquired record is returned as the return value of the query. For example, if the input keyword is “Yokosuka”, “Yokosuka Curry” is acquired from the search log DB 21.

  S2: The keyword search intention candidate specifying unit 12 specifies a word from which the keyword has been deleted from the search query related to the keyword acquired in step S1 as a search intention candidate related to the keyword, and sends it to the search intention classification unit 15. Provide. On the other hand, it is provided to the synonym acquisition unit 13 as a search query pattern in which the word from which the keyword is deleted is replaced with a wild card.

  For example, when the keyword is “Yokosuka” and “Yokosuka curry” is acquired as a related search query by the log filter unit 11 in step S1, the search intention pattern is “curry”, while the search query pattern Is “* curry”. “*” Is a wild card and is a symbol indicating an arbitrary keyword.

  In step S2, the search query acquired by the log filter unit 11 is used. If there are many similar search intentions, the search intention is determined based on the information of the search intention or related text including the search intention. The grouped data may be used for steps S3 and S5 described later.

  S3: Upon receipt of all search query patterns from the keyword search intention candidate specifying unit 12, the synonym acquisition unit 13 refers to the search log DB 21 based on the pattern, acquires a search query that matches the pattern, and performs the search Get and aggregate keywords corresponding to the wildcard part from the query.

  S4: Step S3 is executed for all search query patterns.

  For example, when the search query pattern is “* curry” and there are “Sapporo curry” and “Niigata curry” as related queries, “Sapporo” and “Niigata” are acquired, and the numbers of all patterns are totaled.

  S5: The synonym acquisition unit 13 identifies keywords appearing in a pattern of a predetermined number or a ratio or more as a synonym from the set of keywords acquired in step S4, and provides them to the class search intention candidate identification unit 14.

  S6: The class search intention candidate specifying unit 14 acquires a list of synonyms from the synonym acquisition unit 13, and accesses the log filter unit 11 for each individual synonym to acquire a search query related to the synonym. And the part which extracted the synonym from this search query is acquired as a candidate of the search intention regarding a synonym.

  In step S6, the search query acquired by the log filter unit 11 is used. When there are many similar search intentions, the search intention is determined based on the search intention or related text information including the search intention. The grouped items may be provided to step S8 described later.

  S7: Step S6 is executed for all the isotopes.

  S8: The class search intention candidate specifying unit 14 aggregates the search intentions of all the same terms, and sets the search intention candidates associated with the predetermined number or ratio of the same terms as the search intention candidates related to the class of the same terms. It is specified and provided to the search intention classification unit 15.

  S9: The search intention classification unit 15 acquires the search intention candidate related to the keyword from the keyword search intention candidate specifying unit 12, and acquires the search intention candidate related to the same term of the keyword from the class search intention candidate specifying unit 14. Then, the search intentions included in the search intention candidates related to the same term among the search intention candidates related to the keyword are classified as “general search intentions”. On the other hand, among the search intention candidates related to the keyword, search intentions that are not included in the search intention candidates related to the synonym are classified as “characteristic search intentions”. The classification unit 15 stores these classified search intentions in the search intention DB 22 in association with the keywords.

[Effect of this embodiment]
According to the above-described search intention dictionary creation device 1 of the present embodiment, a search intention dictionary in which search intentions related to a keyword used for a search are classified into more general search intentions and characteristic search intentions is created. . Thereby, for a general search intention, for example, a search using a specialized database can be performed. In addition, special content can be prepared for characteristic search intentions. Therefore, it is possible to examine a strategy for satisfying the search intention, and it is possible to provide information with higher satisfaction to the user.

  In particular, in step S5, words appearing in a predetermined number or a pattern of a predetermined ratio or more are selected as equivalent terms from the set of keywords acquired in step S4, and the equivalent terms can be arbitrarily counted.

  Further, in step S8, search intention candidates associated with a predetermined number or a predetermined ratio of the same term search intention candidates are specified as search intention candidates related to the set of the same term. The search intention candidates can be arbitrarily specified.

[Aspect as Program of the Present Invention]
The present invention can be realized by configuring a program that causes a computer to function as a part or all of the functional units 11 to 15 constituting the search intention dictionary creating apparatus 1 and causing the computer to execute the program. Alternatively, it can be realized by configuring a part or all of the above-described steps S1 to S9 executed by the apparatus 1 with a program that causes a computer to execute the process. The program can be read by a computer-readable recording medium such as FD (Floppy (registered trademark) Disk), MO (Magneto-Optical disk), ROM (Read Only Memory), memory card, CD (Compact). Disk) -ROM, DVD (Digital Versatile Disk) -ROM, CD-R, CD-RW, HDD, SSD, removable disk, etc. can be provided. Alternatively, the program can be provided via the network via the Internet or e-mail.

DESCRIPTION OF SYMBOLS 1 ... Search intention dictionary creation apparatus 11 ... Log filter part (search query extraction means)
12 ... Keyword search intention candidate specifying unit (keyword search intention candidate specifying means)
13 ... Synonym acquisition unit (isotopic acquisition means)
14 ... Class search intention candidate specifying unit (class search intention candidate specifying means)
15 ... Search intention classification section (search intention classification means)
21 ... Search log DB
22 ... Search intention DB (Search intention dictionary)

Claims (8)

A search intention dictionary creation device that creates a dictionary of search intentions of keywords provided for search,
A search query extraction means for extracting a search query including a keyword provided for search and one or more keywords other than the keyword from a search log of a search engine as a search query related to the keyword;
Keyword search intention candidate specifying means for specifying, as a keyword search intention candidate, a word obtained by deleting the keyword from the extracted search query;
For each of the specified search intention candidates, a search query related to the candidate is obtained from the search log, the words from which the search intention candidate portion has been deleted from the search query are totaled, and the words selected from the set of words are allotted. A synonym acquisition means for acquiring words
The obtained synonym is used for the search query extraction means to obtain a search query including the isotope, a word obtained by deleting the isotope from the search query is obtained as a candidate for a search intention of the isotope, and all of the obtained A class search intention candidate specifying means for specifying a set of search intentions selected from candidate search intentions of the same term as a search intention candidate related to the set of peer terms;
Search intention classification means for classifying the keyword into a general search intention and a search candidate specialized for the keyword based on the keyword search intention candidate and the search intention candidate related to the set of the same terms. A search intention dictionary creation device characterized. The search intention classification means classifies search intentions included in the search intention candidates related to the set of the same terms among the search intention candidates of the keywords into general search intentions, and the search intention candidates of the keywords The search intention dictionary creation device according to claim 1, wherein search intentions that are not included in search intention candidates related to a set of synonyms are classified into characteristic search intentions. 3. The search intention dictionary creation device according to claim 1, wherein the synonym acquisition unit selects a word that appears in a pattern of a predetermined number or a predetermined ratio from the set of words as a synonym. 4. . The class search intention candidate specifying means specifies, as a search intention candidate related to the set of isotopes, a search intention candidate that is linked to a predetermined number or a predetermined ratio of the same search word candidates. The search intention dictionary creation device according to any one of claims 1 to 3, wherein: A search intention dictionary creation method executed by a search intention dictionary creation device,
A search query extraction unit of the device extracts a search query including a keyword provided for search and one or more keywords other than the keyword from a search log of a search engine as a search query related to the keyword;
A keyword search intention candidate specifying unit of the device specifying a word obtained by deleting the keyword from the extracted search query as a keyword search intention candidate;
The equivalent word acquisition means of the device acquires a search query related to the candidate from the search log for each of the specified search intention candidates, and totals the words from which the candidate part of the search intention is deleted from the search query. Obtaining a selected word from a set of words as a homolog;
The class search intention candidate specifying means of the device uses the acquired isotope word to the search query extracting means to acquire a search query including the isotope word, and searches for a word obtained by deleting the isotope word from the search query. Acquiring as a candidate for intention, and identifying a set of search intentions selected from candidates of search intentions for all acquired synonyms as search intention candidates related to the set of isotopes; and
A step of classifying the search intention classification means of the device into a general search intention and a search candidate specialized for the keyword based on the search intention candidate of the keyword and the search intention candidate related to the set of the same terms A search intention dictionary creation method characterized by comprising: In the classifying step, search intentions included in search intention candidates related to the set of peer terms among the keyword search intention candidates are classified into general search intentions. 6. The search intention dictionary creation method according to claim 5, wherein search intentions not included in search intention candidates related to a set of words are classified into characteristic search intentions. In the step of identifying as a search intention candidate related to the set of the same terms, the search intention candidates associated with a predetermined number or a predetermined ratio of the same terms are searched for from the acquired search intention candidates of all the same terms. The search intention dictionary creation method according to claim 5 or 6, characterized in that the search intention dictionary is specified as a search intention candidate related to.   A search intention dictionary creation program for causing a computer to function as each means constituting the search intention dictionary creation device according to any one of claims 1 to 4.

Images