JP5068304B2 - Extraction apparatus, method and program - Google Patents

Description

  The present invention relates to an extraction apparatus, method, and program for extracting a query used for content search.

  2. Description of the Related Art Conventionally, a content search system on the Internet uses a method of selecting content by matching with a keyword included in a query based on a query input at a terminal.

  At this time, the content presented as a search result depends on a keyword specified as a query by a user using the search system. That is, whether or not the user can reach the desired content is affected by whether or not an efficient query is input.

  In such a situation, a technique has been proposed in which a search system automatically recommends additional keywords from a database stored in advance and thereby narrows down search results (see, for example, Patent Document 1).

JP 2009-237912 A

  By the way, in recent years, a service for publishing question / answer type content in which an answer is posted from another user in response to a question posted by a certain user has been provided. The contents of such question / answer type content asks / answers, for example, “difference”, “price”, “meaning”, etc. with respect to a certain target (direct object). Therefore, in order to search for these contents, it is effective to simultaneously specify abstract nouns (indirect objects) expressing properties, intentions, types and the like as queries together with direct objects.

  However, in order to classify direct objects and indirect objects and recommend these combinations to the user, most of them are manually maintained by the administrator when preparing the dictionary and recommendation algorithm to be used. I had to rely on it.

  An object of the present invention is to provide an extraction apparatus, method, and program capable of automatically extracting keyword candidates for efficiently searching question / answer type contents.

  The present invention provides the following solutions.

  (1) An extraction device for extracting a query used for searching for question / answer type content, wherein the query is used when any of the content is viewed from a terminal connectable to the extraction device. Together with a query storage means for storing a history of the above, an extraction means for extracting a keyword constituting the query from each of the queries stored in the query storage means, and a keyword extracted by the extraction means, Determining means for determining a category to which each of one or a plurality of keywords co-occurring at a predetermined frequency or more in the same query belongs, and when the variation of the category determined by the determining means is less than a predetermined value, the extracting means If the variation extracted in the category is greater than or equal to a predetermined keyword, Extracting apparatus comprising: a classifying means for classifying the keywords extracted by the means to the second keyword, and a keyword storage means for storing classified the first keyword and the second keyword by the classifying means.

  According to such a configuration, the extraction device stores a history of queries used when any of the question / answer type contents is browsed, and configures the query from each of the stored queries. Extract keywords Then, the extraction device determines a category to which each of one or a plurality of keywords co-occurring in the same query together with the extracted keywords belongs, and if the variation in the categories is less than a predetermined value, If the extracted keyword is the first keyword and the variation of the category is a predetermined value or more, the extracted keyword is classified and stored as the second keyword.

  Therefore, the extraction device can automatically classify the extracted degree of the target keyword as the determination material based on the degree of variation of the category to which the co-occurring keyword belongs. That is, using the fact that the first keyword (direct object) has a small variation (less than a predetermined value) and the second keyword (indirect object) has a large variation (a predetermined value or more), The extraction device can automatically classify direct objects and indirect objects. As a result, the extraction device can automatically extract keyword candidates for efficiently searching for the content from the keywords used when the question / answer type content can be browsed.

  (2) The keyword storage means corresponds to the combination when the keyword co-occurred in the same query is a combination of the first keyword and the second keyword as a result of the classification by the classification means. The extraction device according to (1), which is attached and stored.

  According to such a configuration, the extraction device stores the first keyword (direct object) and the second keyword (indirect object) as a co-occurring combination, and thus has a track record of browsing the content. The query can be used as a recommendation query. As a result, the extraction device can improve the accuracy of the recommendation query.

  (3) It further includes a counting unit that counts the number of times any of the contents has been viewed using a query including the combination, and the keyword storage unit calculates the number of times counted by the counting unit as The extraction device according to (2), which stores the information in association with the combination.

  According to such a configuration, the extraction device browses any of the question / answer type contents by the search based on the combination of the first keyword (direct object) and the second keyword (indirect object). Count the number of times. Therefore, the extraction device can easily select a query that can easily reach the content (a large number of times) by searching by storing the total number of times in association with the combination. As a result, the extraction device can preferentially recommend a query that can easily reach the content.

  (4) The query storage means stores the query used when any of the contents is browsed and the URL of the content in association with each other, refers to the query storage means, and the combination is A selection unit configured to select a content having a relatively high browsing frequency among the content browsed using the included query, wherein the keyword storage unit stores the URL of the content selected by the selection unit; The extraction device according to (2) or (3), which stores the information in association with the combination.

  According to such a configuration, the extraction device stores, as a search history, the query used when any of the question / answer type contents is browsed and the URL of the browsed content in association with each other. To do. In addition, the extraction device selects content having a relatively high browsing frequency from among the browsed content. Therefore, the extraction device stores the URL of the selected content in association with the combination of the first keyword (direct object) and the second keyword (indirect object), thereby making this combination a query. In this case, it is possible to efficiently provide useful content with high browsing frequency.

  (5) When the plurality of first keywords co-occur in the same query as a result of the classification by the classification unit, the keyword storage unit stores the plurality of first keywords in association with each other. The extraction device according to any one of (1) to (4).

  According to such a configuration, the extraction device stores a plurality of first keywords (direct objects) as co-occurring combinations, so that it is meaningful to combine or desired content by combining It is possible to efficiently recommend direct objects that are easy to reach.

  (6) A method in which an extraction device extracts a query used to search for question / answer type content, and is used when any of the content is viewed from a terminal connectable to the extraction device A query storage step for storing a history of the query, an extraction step for extracting a keyword constituting the query from each of the queries stored in the query storage step, and a keyword extracted in the extraction step And a determination step of determining a category to which each of one or a plurality of keywords co-occurring at a predetermined frequency or more in the same query belongs, and the variation of the category determined in the determination step is less than a predetermined value, The keyword extracted in the extraction step as the first keyword, A classification step for classifying the keyword extracted in the extraction step into a second keyword when the variation in the category is equal to or greater than a predetermined value; and a keyword storage for storing the first keyword and the second keyword classified in the classification step And a method comprising:

  According to such a configuration, the same effect as in (1) can be expected when the extraction apparatus executes the method.

  (7) A program that causes the extraction device to execute the method according to (6).

  According to such a configuration, the same effect as in (1) can be expected by causing the extraction apparatus to execute the program.

  According to the present invention, keyword candidates for efficiently searching for question / answer type content can be automatically extracted.

It is a figure which shows the function structure of the extraction server which concerns on 1st Embodiment. It is a figure which shows an example of the query log table stored in query log DB which concerns on 1st Embodiment. It is a figure which shows an example of the direct object candidate table and indirect object candidate table which concern on 1st Embodiment. It is a flowchart which shows the process in the control part which concerns on 1st Embodiment. It is a figure which shows the function structure of the extraction server which concerns on 2nd Embodiment. It is a figure which shows an example of the keyword candidate table which concerns on 2nd Embodiment. It is a flowchart which shows the process in the control part which concerns on 2nd Embodiment.

<First Embodiment>
Hereinafter, a first embodiment, which is an example of an embodiment of the present invention, will be described with reference to the drawings.

[Function configuration]
FIG. 1 is a diagram illustrating a functional configuration of an extraction server 1 (extraction apparatus) according to the present embodiment. The extraction server 1 is an apparatus mainly used by an administrator who manages a system that searches a Web page, particularly a question / answer type content. In other words, the extraction server 1 extracts search keyword candidates that constitute this query for the purpose of algorithm or dictionary data maintenance for determining a query for searching for question / answer type content to be recommended by the search system. To do.

  This embodiment is applied to a computer (extraction server 1) and its peripheral devices. Each unit in the present embodiment is configured by hardware included in a computer and its peripheral devices, and software that controls the hardware.

  In addition to the CPU as the control unit 10, the hardware includes a storage unit 20, a communication unit, a display unit, and an input unit. Examples of the storage unit 20 include a memory (RAM, ROM, etc.), a hard disk drive (HDD), and an optical disk (CD, DVD, etc.) drive. Examples of the communication unit include various wired and wireless interface devices. Examples of the display unit include various displays such as a liquid crystal display and a plasma display. Examples of the input unit include a keyboard and a pointing device (mouse, tracking ball, etc.).

  The software includes a computer program and data for controlling the hardware. The computer program and data are stored in the storage unit 20 and appropriately executed and referenced by the control unit 10. Further, the computer program and data can be distributed via a communication line, and can also be recorded and distributed on a computer-readable medium such as a CD-ROM.

  The control unit 10 of the extraction server 1 includes a keyword extraction unit 11 (extraction unit), a category determination unit 12 (determination unit), and a keyword classification unit 13 (classification unit). The storage unit 20 of the extraction server 1 includes a query log DB (database) 21 (query storage unit) and a keyword candidate DB 22 (keyword storage unit).

  The keyword extraction part 11 extracts the keyword which comprises this query from each of the query memorize | stored in query log DB21. Specifically, the keyword extraction unit 11 delimits the query text with a predetermined delimiter (for example, a space) as a boundary, and extracts each delimited text as a keyword. As a result, a phrase explicitly specified by the user can be used as a keyword. If the query text does not include a delimiter, the keyword extraction unit 11 may perform morphological analysis on the query text and extract a predetermined type of word as a keyword. Good.

  Note that the keyword extraction unit 11 converts a phrase to a representative word by using any synonym or synonym as a representative word for the extracted word or phrase when the synonym or synonym is registered in an existing dictionary. As good as This unifies the variation in expression for each query.

Here, the query log DB 21 stores a history of queries used when any of the question / answer type contents is browsed from a terminal connectable to the extraction server 1.
FIG. 2 is a diagram illustrating an example of a query log table stored in the query log DB 21 according to the present embodiment.

  The query log table stores a query input to various search engines and a URL of content selected (clicked) from a search result list searched by the query. The combination of the query and URL stored in the query log table may be limited to a case where the selected content is a predetermined site (question / answer type content).

  The category determination unit 12 determines a category to which each of one or a plurality of keywords co-occurring at a predetermined frequency or more in the same query belongs together with the keyword extracted by the keyword extraction unit 11. Specifically, when a certain keyword (target word) is extracted, the category determination unit 12 and the keyword (target word) together with another keyword constituting the same query at a predetermined frequency (for example, a ratio). (Co-occurrence word) is extracted. Then, the category determination unit 12 determines to which category the co-occurrence word belongs based on classification rules set in advance for a plurality of categories.

  As a classification rule, for example, a method for determining a category having a relatively high similarity between a feature word and a co-occurrence word provided for each category is used, but the method is not limited to this, and various methods are used. Is applicable.

  When the variation of the category determined by the category determination unit 12 is less than a predetermined value, the keyword classification unit 13 directly classifies the target word extracted by the keyword extraction unit 11 into a target word (first keyword). On the other hand, when the variation of the category is equal to or greater than a predetermined value, the target word extracted by the keyword extraction unit 11 is classified as an indirect object (second keyword).

  The direct object includes, for example, various types (categories) of product names. The indirect object paired with the direct object is a relatively limited type (category) such as “price” or “how to use”. That is, since the category of a keyword that co-occurs with a certain product name has little variation, the keyword classification unit 13 can determine the product name as a direct object. On the other hand, the keyword co-occurring with “price” includes various products and has a large variation in category, so the keyword classification unit 13 can determine “price” as an indirect object.

  The category variation means, for example, the number of determined categories, and the number of categories is relatively small in the number of determined co-occurrence words (for example, the absolute number is less than a predetermined value or the maximum value). The ratio may be counted excluding categories).

Then, the keyword classification unit 13 stores the direct object and the indirect object thus classified in the keyword candidate DB 22.
FIG. 3 is a diagram illustrating an example of a direct object candidate table and an indirect object candidate table stored in the keyword candidate DB 22 according to the present embodiment.

  In the direct object candidate table (a), keywords classified as direct objects by the keyword classifying unit 13 and in the indirect object candidate table (b), keywords classified as indirect objects by the keyword classifying unit 13 Are stored independently.

[Processing flow]
FIG. 4 is a flowchart showing processing in the control unit 10 of the extraction server 1 according to this embodiment. It is assumed that the control unit 10 has sufficiently accumulated a query log, which is a history of queries used when the question / answer type content is browsed, in the query log DB 21 (query storage step).

  In step S <b> 1, the control unit 10 (keyword extraction unit 11) reads a query log that is a history of queries that have been able to reach the question / answer type content from the query log DB 21.

  In step S2 (extraction step), the control unit 10 (keyword extraction unit 11) analyzes each query of the query log read in step S1 and extracts a target word.

  In step S3, the control unit 10 (category determination unit 12) extracts the co-occurrence words that co-occur with the target word extracted in step S2 at a predetermined frequency or more in the query.

  In step S4 (determination step), the control unit 10 (category determination unit 12) determines the category to which each of the co-occurrence words extracted in step S3 belongs.

  In step S5 (classification step), the control unit 10 (keyword classification unit 13) determines whether or not the variation of the category determined in step S4 is less than a predetermined value. When this determination is YES, the control unit 10 moves the process to step S6, and when the determination is NO, the control unit 10 moves the process to step S7.

  In step S6 (classification step), the control unit 10 (keyword classification unit 13) classifies the target word extracted in step S2 directly as a target word because the variation of the category to which the co-occurrence word belongs is small.

  In step S7 (classification step), the control unit 10 (keyword classification unit 13) classifies the target word extracted in step S2 as an indirect object because the variation of the category to which the co-occurrence word belongs is large.

  In step S8 (keyword storage step), the control unit 10 (keyword classifying unit 13) uses the target words classified as direct objects in step S6 and the target words classified as indirect objects in step S7 as keyword candidates. Store as keyword candidates in DB22.

  In step S9, the control unit 10 determines whether or not all the target words have been processed. If this determination is YES, the control unit 10 ends the process. If the determination is NO, the control unit 10 returns the process to step S2 and continues the process for the next target word.

  As described above, according to the present embodiment, the extraction server 1 automatically selects the extracted target word as a direct object or an indirect object using the degree of variation of the category to which the highly co-occurrence keyword belongs as a determination material. Can be classified. Therefore, the extraction server 1 can automatically extract keyword candidates for efficiently searching for the content from the query used when the question / answer type content is browsed.

Second Embodiment
Hereinafter, a second embodiment, which is an example of an embodiment of the present invention, will be described with reference to the drawings. In addition, about the structure similar to 1st Embodiment, the same code | symbol is attached | subjected and description is abbreviate | omitted or simplified.

[Function configuration]
FIG. 5 is a diagram illustrating a functional configuration of the extraction server 1a (extraction apparatus) according to the present embodiment.
In addition to the configuration of the first embodiment, the extraction server 1a further includes a totaling unit 14 and a selection unit 15 in the control unit 10a. Moreover, the information memorize | stored in keyword candidate DB22a of the memory | storage part 20a differs from keyword candidate DB22 of 1st Embodiment.

  When the result of the classification by the keyword classification unit 13 includes a combination of a direct object and an indirect object as a result of the classification by the keyword classification unit 13, the aggregation unit 14 uses the query including this combination to determine any of the question / answer type contents. Count the number of times that is viewed. Specifically, the number of times is the number of times that the user selects (clicks) the question / answer type content from the search result list by the search system.

  As a result of the classification by the keyword classification unit 13, the selection unit 15 refers to the query log DB 21 via the keyword extraction unit 11 when a combination of a direct object and an indirect object is included in the same query, and includes this combination. Content that is viewed using a query that is relatively frequently viewed is selected. The selection unit 15 may select the content having the highest browsing frequency, or may select a plurality of contents having a browsing frequency equal to or higher than a predetermined level.

  The keyword candidate DB 22a associates the combination of the direct object and the indirect object co-occurring in the same query, the number of times counted by the counting unit 14, and the URL of the content selected by the selection unit 15. Remember.

  In addition, as a result of the classification by the keyword classification unit 13, the keyword candidate DB 22a stores a plurality of direct objects in association with each other when a plurality of direct objects co-occur in the same query.

  FIG. 6 is a diagram showing an example of a keyword candidate table stored in the keyword candidate DB 22a according to the present embodiment.

  In the keyword candidate table, the combination of the direct object and the indirect object, the number of times the content is viewed using a query including this combination (the number of clicks), and the combination is frequently used. The content URL is stored in association with each other.

  For example, from a query including “XX”, “□□”, and “Difference” such as “Difference between XX and □□”, the question / answer type content is clicked 1000 times and “http: // ccc It is shown that "~" is browsed most.

[Processing flow]
FIG. 7 is a flowchart showing processing in the control unit 10a of the extraction server 1a according to the present embodiment.

  Steps S11 to S17 are the same as steps S1 to S7 in the first embodiment, and the control unit 10a classifies the target words constituting the query into direct objects or indirect objects.

  In step S18, when the combination of the direct object and the indirect object is included in the same query, the control unit 10a (aggregation unit 14) clicks the question / answer type content from the search result based on this query. Count the number of times.

  In step S19, when the combination of the direct object and the indirect object is included in the same query, the control unit 10a (selection unit 15) selects the URL of the content frequently clicked from the search result based on the query. select.

  In step S20, the control unit 10a (keyword classifying unit 13) tabulates the combinations of the target words classified as direct objects in step S16 and the target words classified as indirect objects in step S17 in step S18. The keyword is stored as a keyword candidate in the keyword candidate DB 22a in association with the number of times and the URL selected in step S19.

  In step S21, the control unit 10a determines whether or not all the target words have been processed. If this determination is YES, the control unit 10a ends the process. If the determination is NO, the control unit 10a returns the process to step S12 and continues the process for the next target word.

  As described above, according to the present embodiment, the extraction server 1a stores the direct object and the indirect object as a co-occurring combination. Accordingly, since the administrator can use a query with a track record of browsing the content as a recommendation query, the extraction server 1a can improve the accuracy of the recommendation query.

  Further, the extraction server 1a stores the number of times the question / answer type content is clicked by the search based on the combination of the direct object and the indirect object. Therefore, since the extraction server 1a can easily select a query that has a large number of clicks and can easily reach the content by the search, an efficient query that can easily reach the content can be preferentially recommended.

  For example, according to the keyword candidate table of FIG. 6, when the keyword “XX” is specified in the search system, the extraction server 1a displays “word of mouth”, “price”, and “difference” in descending order of the number of clicks. A combined query can be recommended.

  Further, the extraction server 1a stores question / answer type content with a high browsing frequency by a search based on a combination of a direct object and an indirect object. Therefore, when this combination is used as a query, the extraction server 1a can efficiently provide useful content with a high browsing frequency.

  Furthermore, since the extraction server 1a stores a plurality of direct objects as co-occurring combinations, it is efficient to generate direct objects that are meaningful for the combination or that can easily reach the desired content by the combination. Can be recommended.

  As mentioned above, although embodiment of this invention was described, this invention is not restricted to embodiment mentioned above. The effects described in the embodiments of the present invention are only the most preferable effects resulting from the present invention, and the effects of the present invention are limited to those described in the embodiments of the present invention. is not.

  In the above-described embodiment, the extraction server 1 or 1a has been described as including each unit. However, the present invention is not limited to this, and each unit may be appropriately distributed to a plurality of servers.

1, 1a Extraction server (extraction device)
10, 10a Control unit 11 Keyword extraction unit (extraction means)
12 Category determination unit (determination means)
13 Keyword classification part (classification means)
14 total part 15 selection part 20, 20a storage part 21 query log DB (query storage means)
22, 22a Keyword candidate DB (keyword storage means)

Claims (7)

An extraction device that extracts a query used to search for question / answer type content,
Query storage means for storing a history of the query used when any of the contents is viewed from a terminal connectable to the extraction device;
Extraction means for extracting a keyword constituting the query from each of the queries stored in the query storage means;
Determining means for determining a category to which each of one or a plurality of keywords co-occurring at a predetermined frequency or more in the same query together with the keywords extracted by the extracting means;
When the variation of the category determined by the determination unit is less than a predetermined value, the keyword extracted by the extraction unit is the first keyword, and when the variation of the category is a predetermined value or more, the keyword is extracted by the extraction unit A classifying means for classifying the keyword into a second keyword;
An extraction apparatus comprising: keyword storage means for storing the first keyword and the second keyword classified by the classification means.   The keyword storage means stores the association of the first keyword and the second keyword in association with each other when the keyword co-occurring in the same query is a combination of the first keyword and the second keyword as a result of the classification by the classification means. The extraction device according to claim 1. A totaling unit that counts the number of times any of the contents is viewed using a query including the combination;
The extraction device according to claim 2, wherein the keyword storage unit stores the number of times counted by the counting unit in association with the combination. The query storage means stores the query used when one of the contents is browsed and the URL of the content in association with each other,
A selection unit that refers to the query storage unit, and selects a content having a relatively high browsing frequency among the content browsed using a query including the combination;
The extraction device according to claim 2, wherein the keyword storage unit stores a URL of the content selected by the selection unit in association with the combination.   The keyword storage means stores the plurality of first keywords in association with each other when a plurality of the first keywords co-occur in the same query as a result of the classification by the classification means. The extraction device according to claim 4. A method in which an extraction device extracts a query used to search for question / answer type content,
A query storage step of storing a history of the query used when any of the contents is viewed from a terminal connectable to the extraction device;
An extraction step of extracting a keyword constituting the query from each of the queries stored in the query storage step;
A determination step of determining a category to which each of one or a plurality of keywords co-occurring at a predetermined frequency or more in the same query together with the keywords extracted in the extraction step,
When the variation of the category determined in the determination step is less than a predetermined value, the keyword extracted in the extraction step is the first keyword, and when the variation of the category is a predetermined value or more, the extraction is performed in the extraction step A classification step of classifying the keyword into a second keyword;
A keyword storage step of storing the first keyword and the second keyword classified in the classification step.   The program which makes the said extraction apparatus perform the method of Claim 6.

Images