JP2011192102A - Device and method for creating summary, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve clarity of a search result screen by reducing useless link display in summary display. <P>SOLUTION: A navigation query and a specific site are identified on the basis of a click log DB70 to obtain an URL of a search result in lower order than the specific site in the URL of a search result searched for in the query. A link URL is obtained from a Web page of the specific site. The obtained URL of the search result is compared with the link URL and a coincident URL is decided as a sub-link to the specific site. Also, the Web page of the sub-link is obtained to generate a title of the sub-link on the basis of a character string in the Web page. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a summary creation device, a summary creation method, and a program.

  Various search services are provided on the Internet as tools for efficiently finding desired information from a vast amount of information. The user performs a search by inputting a query (keyword, term), and browses the web page by an operation such as selecting a URL (Uniform Resource Locator) for browsing the web page from the list of search results.

  On the search result screen displaying the list of search results, the title and summary (summary) of the web page are displayed together with the URL as the search result. The summary includes text that is a summary of the web page and a link (URL) inserted in the web page.

  Further, as a search result display method, a technique is known in which web pages in the same site as a site where a certain web page exists are listed and displayed as the next candidate (for example, see Patent Document 1). By introducing such a display method into the search result summary, a link to a web page of the same domain as each web page can be displayed on the search result screen.

Japanese Patent No. 3539354

  However, a plurality of search results for a query often includes a plurality of URLs of the same domain. When the web page link is displayed in the summary depending on whether or not the domain is the same as in the technique of Patent Document 1, the same duplicate link may be displayed over a plurality of search results. For this reason, useless link display increases and the search result screen is not easily viewed.

  The present invention has been made in view of the above problems, and an object of the present invention is to reduce useless link display in the summary display and improve the visibility of the search result screen.

In order to solve the above problem, in the first invention,
In a summary creation device that creates summary information to be displayed for each of a plurality of ranked link information that is a search result for a query,
Lower link acquisition means for acquiring link information ranked lower than the specific link information in the search result, from the search result;
In-page link acquisition means for acquiring link information included in the page information indicated by the selected link information;
The link information that the link information acquired by the lower link acquisition unit and the link information acquired by the in-page link acquisition unit match is collected as a sublink of the specific link information, and the summary including the sublink is collected. Summary information creation means for creating information;
It is characterized by having.

  According to the first invention, among the search results ranked lower than the specific link information, the link information set in the page information of the link information is collected as a sublink, and the summary information is obtained. In order to create, sublinks are collected so that link information is collected from the lower order of each search result. For this reason, it can prevent duplication about the sublink displayed between each search result. Therefore, it is possible to reduce useless link display and improve the visibility of the search result screen.

  Further, in the second invention, the summary information creating means is configured such that, based on a click log that is a history of individual user selection with respect to the link information, the number of times of user selection among the matching link information is the upper predetermined number. The link information is collected as the sublink.

  According to the second invention, since link information that is frequently selected by the user is collected as a sublink, a sublink that is highly likely to be browsed by the user can be displayed on the search result screen.

  In the third invention, the summary information includes link information of the sublink and a display title, and the summary information creating means is configured to generate the summary information based on a character string included in page information indicated by the sublink. A display title of a sublink is generated.

  According to the third aspect, since the display title is generated from the page information of the sublink, a title representing the content of the sublink can be generated.

  In the fourth invention, the query is selected in the search result for the query based on the click log stored in association with the query and the history of individual user selection for the link information as the search result for the query. It is further characterized by further comprising specific link determining means for determining link information having the highest number of times as the specific link information.

  According to the fourth invention, since the sublink is collected with the link information having the highest selection frequency in the search result as the specific link information, the sublink can be created for the main link information in the search result. .

  According to the present invention, as a sublink in the summary information displayed in each search result, duplication between each search result is prevented and useless link display is reduced, thereby making it easy to view the search result screen. Can be improved.

The figure which shows the structural example of the system concerning one Embodiment of this invention. The figure which shows the data structure example of click log DB. The figure which shows the data structure example of link information DB. The flowchart which shows the process example of a sublink creation process. The flowchart which shows the process example of a navigational query extraction process. The flowchart which shows the process example of a sublink title production | generation process. The figure which shows the example of a display of a search result screen.

  Hereinafter, preferred embodiments of the present invention will be described. FIG. 1 is a block diagram showing an example of a system configuration when the summary creation device of the present invention is applied to a search server 1.

〔System configuration〕
According to FIG. 1, the search server 1 is connected to a user terminal T via a network such as the Internet, and performs data communication with the user terminal T according to the HTTP protocol or the like.

[User terminal]
The user terminal T is realized by software (program) executed on computer hardware such as a personal computer. As shown in FIG. 1, the user terminal T is input by an input unit 100 for inputting a query, a search execution instruction, and the like. Request transmission unit 110 for transmitting a request such as a search request corresponding to the received query or a web page browsing request, a response reception unit 120 for receiving a response transmitted from the server in response to the request, and a web received together with the response And a display unit 130 for displaying the page on a display or the like. The request transmission unit 110 and the response reception unit 120 can be realized by a so-called web browser.

[Search Server]
The search server 1 receives a query from the user terminal T, executes a search process, and creates / transmits a web page for displaying a search result.

  As shown in FIG. 1, the search server 1 includes a request receiving unit 10, a search engine 20, a page generating unit 40, a search result transmitting unit 50, a navigational query extracting unit 60, a click log DB 70, a navigational query DB 80, and a search. An index 90 is provided.

  The request receiving unit 10 is a functional unit that receives a request such as a search request including a query or a transmission request for a web page from the user terminal T, and includes a query acquisition unit 12 and a redirect processing unit 14. .

  When the received request is a search request, the query acquisition unit 12 acquires a query included in the search request and transmits it to the search engine 20.

  The search engine 20 searches the search index 90 based on the query, and outputs a list of URLs that are the search results. The search index 90 is a database in which web pages on the Internet are indexed and stored so as to be searchable. As shown in FIG. 3, the index keyword, URL, rank, main title, and summary text are stored. And the sublink title are stored in association with each other.

  The main title is display information (character string) corresponding to the URL of the search result, and is displayed instead of the URL. The summary text is content describing an outline of the web page for each search result. In the display of the search result, as shown in FIG. 7, a link to the URL is set in the main title, and summary text (for example, summary SM) is displayed in association with it.

  In addition, when a link is included in the web page indicated by the URL of the search result, the link is displayed as a sublink (for example, sublink SL in FIG. 7). The sublink title of the search index 90 is display information (character string) for display instead of the URL to the sublink.

  The search engine 20 includes a crawler (not shown) that collects web pages from the Internet, an indexer (not shown) that indexes the web pages so as to be searchable, and a sublink title generation unit 22.

  The crawler collects web pages that exist on the Internet by going around other web pages linked to each web page.

  The indexer performs language processing or the like on the web pages collected by the crawler to generate an index (index), and stores and stores the URL (link information) of the web page in association with the index. In the case of this index, ranks are given by various ranking algorithms using the importance of keywords in a web page, weights based on link relationships between pages, and the like.

  Further, the search engine 20 extracts a character string as a title from a web page corresponding to a URL displayed as a search result, and creates a main title. Specifically, a web page is acquired, a title portion (for example, a character string enclosed by title tags) in the web page is extracted, and the main title is associated with the URL.

  In addition, text that can be summary text is extracted from the web page corresponding to the URL of each search result and set as summary text. For example, a feature word may be extracted by performing language processing on a web page, a predetermined number of character strings including the feature word may be extracted as summary text, or a character string to be displayed in the web page may be extracted. A predetermined number of character strings may be extracted from the beginning.

  The sublink title generation unit 22 of the search engine 20 generates a sublink title based on a sublink title generation process (see FIG. 6) described later.

  In response to a search request from the search server 1, the search engine 20 reads URLs associated with indexes that match the query, ranks the search results, generates a list, and returns the list to the search server 1. .

  The redirect processing unit 14 receives a redirect URL access from the user terminal T and performs a redirect process for returning a redirect response. The redirect process receives access to the search server 1 based on the redirect URL embedded in a web page such as a list of search results, and switches the access to the original URL of the target site included in the redirect URL. Is a process of transmitting a redirect response to the user terminal T.

  When the response reception unit 120 of the user terminal T receives the redirect response from the search server 1, the request transmission unit 110 accesses the website indicated by the URL based on this response via the Internet N to request a web page. Send.

  The search server 1 generates a redirect URL including, as parameters, a URL of the search result, a URL indicating the search server 1, and a query, and embeds the redirect URL corresponding to each search result in the web page. By transmitting this web page to the user terminal T, a list of search results is displayed on the user terminal T.

  Thus, when the user performs a selection operation for browsing the search result, the search server 1 is first accessed from the user terminal T. Then, the user terminal T accesses the website corresponding to the search result by the redirect response returned by the search server 1, and displays the web page corresponding to the URL.

  The redirect processing unit 14 includes a click log collection unit 16. The click log collection unit 16 acquires the combination of the query and the URL of the search result from the redirect URL selected (clicked) at the user terminal T at the timing when the redirect process is performed, and records it in the click log DB 70 as a click log. . By collecting this click log, the frequency of clicks on search results can be calculated.

  The click log DB 70 is a database for accumulating and storing click logs acquired by the click log collection unit 16, and stores a query, a click URL, and the number of clicks in association with each other as shown in FIG.

  That is, the click log collection unit 16 stores in the click log DB 70 a combination of a query searched by the user through a web search and a URL (click URL) selected in the search result, and clicks corresponding to the combination. Count and update the number of times.

  The page generation unit 40 obtains a search result that is a list of ranked URLs from the search engine 20 and generates a web page for displaying the search result. The page generation unit 40 includes a search result acquisition unit 42 and a sublink creation unit 44.

  The search result acquisition unit 42 acquires a list of URLs that are search results for the query from the search engine 20. At this time, the main title, summary text, and sublink title associated with the URL of the search result are also acquired.

  The sublink creation unit 44 creates a sublink to be displayed corresponding to the URL acquired by the search result acquisition unit 42 based on a sublink creation process (FIG. 4) described later.

  When the page generation unit 40 generates a web page for displaying the search result (hereinafter referred to as “search result page”), the page generation unit 40 is associated with the URL of the search result acquired by the search result acquisition unit 42. The URL is set as a title link destination and included in the search result page.

  Further, summary information associated with the URL of the search result is read out and included in the search result page in association with the URL. At this time, the URL of the sublink is set as the link destination of the sublink title corresponding to the URL of the sublink created by the sublink creating unit 44. A display example of the search result page generated in this way is shown in FIG. As shown in FIG. 7, the summary SM and the sublink SL associated with each search result are displayed on the search result screen W displayed based on the search result page.

  The search result transmission unit 50 transmits the search result page generated by the page generation unit 40 to the user terminal T from which the request reception unit 10 has received the search request.

[Navigation Query]
In the present embodiment, the sublink creation unit 44 creates a sublink for each search result for a specific site (URL) having a navigational query.

  The navigational query is a query that is searched for the purpose of accessing (moving) to a specific site. For example, there are many company names, site names, university names, personal names, and the like. The navigational query extraction unit 60 extracts a combination of a navigational query and a URL serving as a specific site corresponding to the query based on the click log DB 70.

  Based on the navigational query extraction process shown in FIG. 5, the navigational query extraction unit 60 calculates a first score, which is an index indicating the probability as a navigational query based on the number of clicks for each query, The second score, which is an index indicating the probability of the navigational query based on the click rate, is calculated every time, and the combination of the query and URL extracted based on the score for each query is stored in the navigational query DB 80. .

  The navigational query DB 80 (not shown) is a database that stores a navigational query and a URL of a specific site corresponding to the navigational query in association with each other.

  Hereinafter, the navigational query extraction process is performed for each query in the click log DB 70 by a batch process or the like at a predetermined timing. Specific processing will be described below.

  First, the navigational query extraction unit 60 selects one query from the click log DB 70 (step S21), and the largest number of clicks of the first to Nth clicks of the click URL associated with the query. A thing is acquired as a variable MAX1 (step S22).

  Next, the average value of the number of clicks is set in the variable AVE1 (step S23), and the number of clicks corresponding to the first to Nth URLs except the one with the largest number of clicks is set as the variable EACH1 in the calculation of Σ. The first score is calculated by the following formula (step S24).

[Formula 1]
First score = (MAX1-AVE1) ² −Σ (EACH1-AVE1) ²

  Next, the navigational query extraction unit 60 calculates the maximum click frequency rate of the first to Nth URLs as the variable MAX2 (step S25), and sets the average value of the click frequency rate to the variable AVE2. Set (step S26). The click frequency rate is the ratio of the number of clicks of each URL to the total value of the clicks of all URLs associated with the query.

  Then, the second score is calculated by the following equation using the click frequency rate of each of the first to Nth URLs except the one with the maximum click frequency rate as a variable EACH2 in the calculation of Σ (step S27).

[Formula 2]
Second score = (MAX2-AVE2) ² −Σ (EACH2-AVE2) ²

  Next, the navigational query extraction unit 60 determines whether or not each of the first score and the second score is equal to or greater than a predetermined threshold, and if each of the first score and the second score is equal to or greater than the threshold, the step The query to be processed selected in S21 and the click URL having the maximum number of clicks (rate) are associated with each other and stored in the navigational query DB 80 (step S28).

  By performing the above series of processes for all queries stored in the click log DB 70, if each query is a navigational query, the corresponding query URL (specific site) and the corresponding navigation URL are stored in the navigational query DB 80. Remembered.

  As described above, in the navigational query extraction process, click log data in the search service is analyzed to automatically select a navigational query and a non-navigational query. At this time, by applying a method of calculating the variance value, the first score and the second score, which are indexes indicating the probability as the navigational query, are calculated, and those whose both exceed a predetermined threshold are used as the navigational query. Extracting.

  Here, in the case of a navigational query, there is a characteristic that one site in the search list is clicked in a concentrated manner. Therefore, the first score and the second score are used for determining whether the site is a navigational query. Is used. If one element is prominent compared to the other elements, the variance value is high. Therefore, in the case of a navigational query, the variance value is high. Moreover, if all the elements are close to the average value, the variance value becomes small.

  However, if each element varies, the deviation from the average value increases and the variance value also increases. For this reason, even when a portion that is not clicked is divided into several locations, the variance value becomes high, and it is not possible to accurately determine whether or not it is a navigational query.

  Therefore, the variance value is not applied as it is, but the deviation square sum of other clicks is subtracted from the deviation square of the maximum clicks. As a result, in the case of a navigational query where a specific site is clicked in a concentrated manner, the value of deviation squared by the number of clicks increases, and the value of the sum of deviation squared of other clicks becomes relatively small. The overall score value increases.

  In addition, when the clicked part is not divided into several parts (in the case of not being a navigational query), the deviation sum of squares of other clicks becomes large, and the score value is lowered. The advantage of this method is that the difference becomes smaller when the number of clicks varies, and the difference value becomes a negative value when clicked equally. As a result, it can be used as an index indicating a state where only one site is clicked in a concentrated manner.

  The second score is calculated based on the click frequency rate, whereas the first score is calculated based on the click frequency. That is, since the score of a query with a high search request can be high, the influence is removed by the second score.

  For example, when 1% of the number of searches is clicked, a query that is popular because it is a popular site, and a query that is not frequent (low search) When scoring is performed only by the number of clicks, the score of a query with a higher number of searches is higher.

  Therefore, scoring is performed on the shape of the click dispersion by calculating the second score by replacing the click frequency with the click frequency rate in the same formula as the first score calculation formula. However, when calculating based on the click frequency rate, the score of a query with low exposure will be high, so it is not used alone, but a second score is predetermined for those whose first score is greater than a predetermined threshold. It is used to confirm whether or not it is larger than the threshold value. Thereby, the determination accuracy of the navigational query can be increased.

  As described above, the specific site corresponding to the navigational query is extracted, and the specific site is set as a sublink creation target. This makes it possible to display sublinks for specific sites that are frequently accessed for queries rather than creating sublinks for all URLs.

[Sublink creation processing]
Next, sublink creation processing by the sublink creation unit 44 will be described with reference to FIG.

  First, the sublink creation unit 44 identifies a URL as a sublink creation target (step S11). In the present embodiment, when the query acquired by the query acquisition unit 12 is a navigational query, the URL associated with the query in the navigational query DB 80 described above is specified as a creation target.

  Then, the web page indicated by the specified URL is acquired from the Internet, and the link URL set by the href tag or the like in the web page is extracted (step S12).

  In addition, a search result URL (search URL) is acquired from the search engine 20 from the search result acquisition unit 42 (step S13), and a search URL lower than the URL of the sublink creation target is selected from the search URL. Selected (step S14).

  The sublink creation unit 44 compares the link URL extracted in step S11 with the search URL selected in step S14, and collects URLs determined to match (step S15). That is, URLs that are ranked lower than the URL to be created are collected from the URLs searched by the navigational query.

  Then, the number of clicks associated with the combination of each collected link and the query acquired from the query acquisition unit 12 is acquired from the click log DB 70, and a predetermined number (for example, 3) with the highest number of clicks in the link is acquired. Is determined as a sublink (step S16).

  Then, a sublink title corresponding to the sublink determined for the URL to be created is acquired from the sublink titles acquired by the search result acquisition unit 42 (step S17). Note that the sublink creation unit 44 may directly access the search index 90 to obtain the sublink title associated with the sublink URL determined in step S16.

  The sublink creation unit 44 adds a sublink to be displayed in the summary of the URL to be created, and generates a search result page (step S18).

[Sublink title generation processing]
Next, sublink title generation processing performed by the sublink title creation unit 22 will be described with reference to FIG.

  First, the web page indicated by the URL stored in the search index 90 is acquired (step S31). Then, the query associated with the URL, that is, the query input to reach the URL is read from the click log DB 70 (step S32).

  Then, the character string in the read query is compared with the character string in the web page, and a similar character string is extracted from the web page (step S33). Specifically, the character string included in the query and the character string in the web page are divided by N-gram or the like, and the similarity is calculated by the number of matching characters in the divided character string. Extract strings in queries with high similarity. The character string extracted in this way is stored in the search index 90 as a sublink title corresponding to the URL (step S34).

  For example, as shown in FIG. 3, the web page of URL “http://www.yaho.co.jp/corp” is acquired. Then, “Yahoo” and “Yahoo company information” illustrated in FIG. 2 are read as queries associated with this URL, and character strings similar to the character strings such as “Yahoo” and “Company information” included in this query are read. Extract from "http://www.yaho.co.jp/corp" web page.

  At this time, for example, when both “Yahoo” and “Company information” are extracted as character strings having a high degree of similarity, the concatenated character strings may be used as sublink titles, but they are associated with one URL. When there is an overlapping character string between a plurality of sublink titles and the main title, the character string at the overlapping portion may be deleted. Thereby, a sublink title can be generated based on a query used when a user accesses a sublink.

  The sublink title generation process may be as follows. That is, a web page indicated by a URL is acquired, and morphological analysis is performed on the web page to extract words with high appearance frequency and characteristic words based on TF / IDF, and the words are sub-linked titles. Thereby, a sublink comes to be represented by the word showing the characteristic of a web page. Further, since it is desirable that the sublink title displayed together with the search result is expressed by a short word, a simple summary can be created by generating a title in units of words as described above.

  A display example of the search result screen W generated using the summary information created as described above is shown in FIG. According to FIG. 7, the summary SM is displayed for each URL of the search result in the search result screen.

  Further, the sublink SL is displayed for the “Yahoo home page” specified as the site stored for the navigational query “Yahoo”. The sublink SL has a high click count among URLs (search results) ranked lower than the “Yahoo homepage” in the search results.

  In this way, even in the sublink displayed as a summary, URLs that are ranked lower than the search results to be displayed and that are frequently viewed by the user are collected and displayed for the search results. Become so. Therefore, in the summary display of the search results, it is possible to reduce the useless link display without performing the duplicate link display between the search results, and to improve the visibility of the search result screen.

<Modification>
The present invention has been described above by the preferred embodiments of the present invention. While the invention has been described with reference to specific embodiments, various modifications and changes may be made to the embodiments without departing from the broad spirit and scope of the invention as defined in the claims. Obviously you can. In other words, the present invention should not be construed as being limited by the details of the specific examples and the accompanying drawings.

  For example, the target URL for creating a sublink is specified as a site for a navigational query, but a URL with a high number of clicks may be specified for the query, and a sublink may be created for the URL. The URL may be created for each search result URL. Also in this case, since the sublinks are created so that the URLs of the search results ranked below each search result are collected, it is possible to prevent display of duplicate sublinks.

DESCRIPTION OF SYMBOLS 1 Search server 10 Request receiving part 12 Query acquisition part 14 Redirect process part 16 Click log collection part 20 Search engine 40 Page generation part 42 Search result acquisition part 44 Sublink creation part 50 Search result transmission part 60 Navigational query extraction part 70 Click Log DB
80 Navigational Query DB
90 Search index SL Sublink SM Summary T User terminal

Claims (6)

In a summary creation device that creates summary information to be displayed for each of a plurality of ranked link information that is a search result for a query,
Lower link acquisition means for acquiring link information ranked lower than the specific link information in the search result, from the search result;
In-page link acquisition means for acquiring link information included in the page information indicated by the selected link information;
The link information that the link information acquired by the lower link acquisition unit and the link information acquired by the in-page link acquisition unit match is collected as a sublink of the specific link information, and the summary including the sublink is collected. Summary information creation means for creating information;
A summary creation device comprising: The summary information creating means includes:
Based on a click log that is a history of individual user selection with respect to the link information, among the matching link information, link information having a predetermined number of higher user selections is collected as the sublink. The summary creation apparatus according to claim 1. The summary information includes link information and a display title of the sublink,
The summary information creating means includes:
The summary creation device according to claim 1, wherein a display title of the sublink is generated based on a character string included in page information indicated by the sublink. Based on the click log stored in association with the query and the history of individual user selection for the link information as a search result for the query,
The summary creation device according to any one of claims 1 to 3, further comprising specific link determination means for determining, as the specific link information, link information having the highest number of selections in a search result for the query. . In a summary creation method in which a computer creates summary information to be displayed for each of a plurality of ranked link information that is a search result for a query,
A lower link acquisition step of acquiring link information ranked lower than the specific link information in the search result from the search result;
In-page link acquisition step of acquiring link information included in the page information indicated by the selected link information;
Collecting link information in which the link information acquired in the lower link acquisition step matches the link information acquired in the intra-page link acquisition step as a sublink of the specific link information, and the summary including the sublink Summary information creation step to create information,
And a summary creation method. A program for causing a computer of a summary creation device to create summary information to be displayed for each of a plurality of ranked link information, which is a search result for a query,
A lower link acquisition step of acquiring link information ranked lower than the specific link information in the search result from the search result;
In-page link acquisition step of acquiring link information included in the page information indicated by the selected link information;
Collecting link information in which the link information acquired in the lower link acquisition step matches the link information acquired in the intra-page link acquisition step as a sublink of the specific link information, and the summary including the sublink Summary information creation step to create information,
For causing the computer to execute.

Images