JP2009116422A - Query extraction method, query extractor, and query extraction program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a query extraction method, a query extraction device and a query extraction program for extracting information of the corresponding relation of a site corresponding to a navigation query with sufficient precision. <P>SOLUTION: This query extraction method includes: a process for summing up click logs showing the history of each selection to a plurality of sites from a retrieval list to the query by each query; a process for calculating a first distribution value based on the click frequency of each site corresponding to each query on the basis of the clock log; and a process for extracting the query whose calculated first distribution value is equal to or more than a prescribed value and site information whose click frequency in the retrieval list for the query is the highest. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a search technique in an Internet portal site or the like.

  Since an enormous amount of information exists on the Internet, various search services are provided on portal sites and the like as tools for efficiently finding desired information from the information. The user performs a search by inputting a query (keyword, term), and browses the site by a search operation such as selecting a desired site from a list of search results.

  By the way, the search service is often used not only for searching a page (article) on which information desired to be found but also for the purpose of moving to a specific site. For example, when a company name, a site name, a university name, an individual name, etc. are input, a site operated by the subject is searched. A query input in such a case is called a navigational query.

  In addition, the web browser has a bookmark function in which the URL of a desired web page is registered in advance by the user. Registration to the bookmark is generally performed by the user performing a menu operation for bookmark registration in a state where a user displays a desired web page on a web browser.

  However, since the user browses a plurality of sites one after another, when registering in a bookmark after transition to another web page, it is necessary to perform a registration operation after displaying a desired web page again, The operation is complicated. In particular, since the web page registered in the bookmark is often the top page of the site, it takes time to trace the browsing history of the user.

  Therefore, a search site may be used instead of a bookmark. That is, the user selects and inputs a query that outputs a desired site as a search result at the search site, and browses the desired site from the search result. From this point of view, the navigational query can be rephrased as a query input for browsing a site that the user has not bookmarked.

It is effective in improving the search service that the portal site providing the search service can extract such navigational queries and information on the correspondence between the sites corresponding to the navigational queries. For example, when a search is performed using a navigational query, the site corresponding to the navigational query can be brought to the top of the search result, or the user can be guided to the site corresponding to the navigational query. It becomes like this.
JP 2001-31518 A

  As described above, it is useful to be able to extract information on the correspondence between sites corresponding to navigational queries, but there has been no effective method for this purpose. That is, as a method for extracting the navigational query, a method of statically analyzing the URL (Uniform Resource Locator) of the site, the title of the site, etc. can be considered, but the site corresponding to the navigational query varies with time. There is a possibility, and the static analysis method as described above is not sufficient. For example, there are official websites and blog sites of a celebrity. Normally, many users searched by celebrity name queries and accessed the official website. May come to access the site. Thus, the navigational query may vary depending on the time.

  On the other hand, in Patent Document 1, a search is performed using a database stored so that one URL corresponds to a predetermined keyword, and the number of accesses to a link from the search result is counted. A search system that can count the needs for a computer is disclosed. However, it does not take into account the above-mentioned properties of navigational queries, and cannot be used as a method for extracting navigational queries.

  The present invention has been proposed in view of the above-described conventional problems, and an object of the present invention is to provide a query extraction method capable of extracting information on correspondence between a navigational query and a corresponding site with sufficient accuracy. Another object of the present invention is to provide a query extraction device and a query extraction program.

  In order to solve the above problems, according to the present invention, as described in claim 1, a click log indicating a history of individual selections for a plurality of sites from a search list for a query is aggregated for each query. A step of calculating a first variance value based on the number of clicks per site for each query based on the click log, a query in which the calculated first variance value is a predetermined value or more, and The gist is a query extraction method including a step of extracting site information having the highest number of clicks in a search list for a query.

  In addition, as described in claim 2, a step of counting a click log indicating a history of individual selections for a plurality of sites from a search list for a query for each query, and each query based on the click log Calculating a first variance value based on the number of clicks per site with respect to the site, and calculating a second variance value based on the click frequency rate per site for each query based on the click log. A query extraction method comprising: a query in which the first variance value and the second variance value are each equal to or greater than a predetermined value; and a step of extracting site information having the highest number of clicks in a search list for the query. can do.

  Moreover, as described in claim 3, in the query extraction method according to claim 1 or 2, the step of calculating the first variance value is based on a deviation square of the maximum number of clicks. The first variance value can be calculated by subtracting the deviation sum of squares of other clicks.

  In addition, as described in claim 4, in the query extraction method according to claim 3, a button for changing the search list is displayed on the screen displaying the search list for the query, The step of calculating the variance value of 1 can further calculate the first variance value by subtracting a deviation square of the number of clicks of the displayed button.

  In addition, as described in claim 5, in the query extraction method according to claim 2, the step of calculating the second variance value includes the step of calculating the second click value rate from the square of the deviation of the maximum click number rate. The second variance value can be calculated by subtracting the deviation sum of squares.

  Further, as described in claim 6, in the query extraction method according to claim 5, a search list for the query and a button for changing the displayed search list are displayed on one screen, The step of calculating the second variance value may further calculate the second variance value by subtracting a deviation square of the click frequency rate of the displayed button.

  Moreover, as described in claim 7, in the query extraction method according to any one of claims 1 to 6, the step of counting the click logs includes counting the click logs for each user, In the step of calculating the variance value, the click log corresponding to the query input by the user for the second time or later can be used for the calculation of the variance value based on the click log for each user.

  Further, as described in claims 8 and 9, it can be configured as a query extraction device.

  Further, as described in claims 10 and 11, it can be configured as a query extraction program.

  In the query extraction method, the query extraction device, and the query extraction program of the present invention, a variance value based on the number of clicks for each site for each query is calculated based on the click log of the search service, and the variance value is equal to or greater than a predetermined value. And the site information with the highest number of clicks in the search list for the query is extracted, so the information on the correspondence between the navigational query and the corresponding site can be extracted with sufficient accuracy. .

  Hereinafter, preferred embodiments of the present invention will be described.

<System configuration>
FIG. 1 is a diagram showing a configuration example of a system according to an embodiment of the present invention.

  In FIG. 1, a network 2 such as the Internet includes a plurality of client terminals 1 used by a user, a plurality of Web servers 4 that provide contents to be searched and viewed, and a search server 3 to which the present invention is applied. Is connected.

  The client terminal 1 has a query input unit 101 that inputs a query from the user, a search execution instruction reception unit 102 that receives a search execution instruction from the user, and the search server 3 each time a query is input by the query input unit 101. In addition to transmitting a query, the search execution instruction receiving unit 102 includes a query transmission unit 103 that transmits a query to the search server 3 when a search execution instruction is received.

  The client terminal 1 also receives a recommended site (summary and URL of a site particularly recommended as a navigational query) from the search server 3, and a search list from the search server 3 (a list of sites hit by the search). A search list receiving unit 105 that receives a summary and a list of URLs), a display control unit 106 that controls screen display based on the recommended site received by the recommended site receiving unit 104 and the search list received by the search list receiving unit 105; It has. Note that the URL included in the recommended site and the search list is acquired on the search server 3 side as a click log of the site selection history by the user from the search result. The URL is a redirect URL that is accessed and redirected from there. The redirect URL includes a URL used as an entry to the search server 3 and information specifying the query used for the search and the selected site.

  The client terminal 1 also includes a site selection receiving unit 107 that receives a site selection by the user from a recommended site or a search list displayed by the display control unit 106, and a URL (redirect URL) of the site selected by the site selection receiving unit 107. ) To the search server 3, receives a redirect response from the search server 3 to instruct to switch access to the URL of the target site, and accesses the Web server 4 that is the target site; It has.

  The function units 101 to 108 in the client terminal 1 are realized by computer software (program) executed on computer hardware.

  On the other hand, the search server 3 receives a query from the client terminal 1 via the network 2 (a query that is simply input by a user or a search execution instruction may be accompanied), Based on the query received by the query receiving unit 301, in the case of a query that is simply input by the user, the navigational query DB 312 is searched, and when a navigational query is found, it is output as a recommended site and the search is executed. In the case of a query with an instruction, a search processing unit 302 that searches the content DB 311 and outputs a search list is provided.

  In addition, the search server 3 transmits a recommended site output from the search processing unit 302 to the client terminal 1 via the network 2 and a search list output from the search processing unit 302 to the network 2. A search list transmission unit 304 that transmits to the client terminal 1.

  Further, the search server 3 accepts an access by the redirect URL from the client terminal 1 via the network 2, and sends a query from the redirect URL at the timing when the redirect processing unit 305 performs processing by the redirect processing unit 305 that returns a redirect response. And a click log collection unit 306 that identifies the selected site and records the click log in the query log DB 313.

  Further, the search server 3 calculates a first score (first variance value) that is an index indicating the probability as a navigational query based on the number of clicks for each query based on the query log in the query log DB 313 at a predetermined timing. Then, the first score calculation unit 307 registered in the scoring DB 314 and the second score (second variance value) that is an index indicating the probability as a navigational query based on the click frequency rate for each query are calculated and scored. A second score calculation unit 308 registered in the ring DB 314 and a navigational query extraction unit 309 that extracts a navigational query based on the scoring result of the scoring DB 314 and registers it in the navigational query DB 312 are provided.

  The functional units 301 to 309 in the search server 3 are realized by computer software (program) executed on computer hardware.

  FIG. 2 is a diagram showing an example of the data structure of each database provided in the search server 3.

  FIG. 2A shows an example of the data structure of the content DB 311, which includes a “query” field and a “site URL” field corresponding to the query. The “site URL” field may include a plurality of URLs.

  FIG. 2B shows an example of the data structure of the navigational query DB 312, and includes a “navigation query” field and a “site URL” field corresponding to the navigational query. In principle, there is one site URL corresponding to the navigational query.

  FIG. 2C shows an example of the data structure of the query log DB 313. The “query URL” field, the “first URL and click count” field, and “first record” corresponding to the display order of the search list for the query are shown. URL click frequency rate field,..., “Nth URL and click frequency” field, “Nth URL click frequency rate” field, “total click frequency” field, and “average click frequency” field And an “average click frequency rate” field, a “next / research” click frequency field, and a “next / research” click frequency rate field. Here, the number of clicks of the 1st to Nth URLs is an integrated value within a predetermined period of the number of times a site corresponding to the URL of the rank is selected by a plurality of users from the search list for the query. The total number of clicks is the total number of clicks of the 1st to Nth URLs for the query. The average number of clicks is an average value of the number of clicks of the 1st to Nth URLs for the query. The number of clicks on “next / re-search” is the number of times that the user did not select a site from the N search lists initially displayed at the time of search and switched to the next search list page or re-searched. It is. Further, “˜frequency ratio” is a ratio in which the sum of the total number of clicks of all the queries to be aggregated is used as a parameter.

  Note that the first to Nth URLs corresponding to the search list display order may be managed in a separate table. Also, the click frequency rate of the 1st to Nth URLs, the total click frequency, the average click frequency, the average click frequency rate, the “next / re-search” click frequency rate, etc. are not included in the table, and are calculated at the time of calculating the score described later. You may do it.

  FIG. 2D shows an example of the data structure of the scoring DB 314. The “query” field, the “first score” field for the query, the “second score” field, and the corresponding “maximum number of clicks ( Rate) URL "field. The maximum number of clicks (rate) URL is a URL corresponding to a site that is recognized as the maximum number of clicks in the first / second score calculation. Note that the maximum number of clicks (rate) URL may be managed in a separate table.

<Operation>
FIG. 3 is a sequence diagram illustrating a processing example of the above-described embodiment.

  In FIG. 3, when a user inputs a query to the client terminal 1 using the query input unit 101 (step S101), the query transmission unit 103 of the client terminal 1 transmits a query without a search execution instruction to the search server 3. (Step S102).

  When the query receiving unit 301 receives a query not accompanied by a search execution instruction, the search server 3 searches the navigational query DB 312 using the search processing unit 302 (step S103). At this time, when the input query exists in the navigational query DB 312, the search processing unit 302 outputs a site corresponding to the navigational query as a recommended site. Then, the recommended site transmission unit 303 of the search server 3 transmits the recommended site, which is the search result of the search processing unit 302, to the client terminal 1 (step S104).

  When the recommended site is received by the recommended site receiving unit 104, the client terminal 1 displays the recommended site on the screen by the display control unit 106 and presents it to the user (step S105). FIG. 4A shows an example of a search screen, and shows a state where a balloon 12 indicating a recommended site is displayed as a result of the query “Taro” being entered in the query input field 11. In this case, the user can access and browse the site by clicking the OK button 13.

  Returning to FIG. 3, when the user issues a search execution instruction and is received by the search execution instruction reception unit 102 (step S106), the query transmission unit 103 of the client terminal 1 accompanies the search server 3 with the search execution instruction. A query is transmitted (step S107).

  When the query receiving unit 301 receives a query with a search execution instruction, the search server 3 searches the content DB 311 by using the search processing unit 302 and outputs a search list (step S108). Here, it is assumed that the search processing unit 302 performs normal search processing using the content DB 311. However, when the navigation query DB 312 is searched together with the content DB 311 and the input query is a navigation query (navigation query). If there is a search hit in the null query DB 312), the display order may be changed so that the site corresponding to the navigational query is displayed at the top of the search list.

  Next, the search list transmission unit 304 of the search server 3 transmits the search list that is the search result of the search processing unit 302 to the client terminal 1 (step S109).

  When the search list receiving unit 105 receives the search list, the client terminal 1 causes the display control unit 106 to display the search list on the screen and present it to the user (step S110). FIG. 4B shows an example of a search screen. As a result of the query “XX Taro” being entered in the query input field 11 and the search execution instruction button 14 being pressed, the search list 15 is “1. Blog 2. XXX Taro Official Website 3. XXX Taro Fan Club ... "is displayed. A search list and other page selection field 16 is also displayed at the bottom of the search screen. Clicking “Next” displays the next page of the search list, and clicking a page number allows any page to be displayed. A search list of is displayed.

  Returning to FIG. 3, the user selects a desired site from the search list or recommended site display, and when this is received by the site selection receiving unit 107 (step S111), the redirect processing unit 108 of the client terminal 1 searches the search list or recommended site. The search server 3 is accessed based on the redirect URL embedded in the site (step S112). When receiving the access, the redirect processing unit 305 of the search server 3 transmits a redirect response instructing switching of access to the original URL of the target site to the client terminal 1 (step S113). In parallel with this, the click log collection unit 306 of the search server 3 identifies the query and the selected site from the redirect URL, and records the click log in the query log DB 313 (step S114). The click log collection unit 306 also stores the click log in the query log DB 313 when the client terminal 1 requests the search server 3 to display another page of the search list or when a re-search is requested. Record.

  Then, the redirect processing unit 108 of the client terminal 1 accesses the Web server 4 based on the redirect response (Step S115), and the Web server 4 transmits a response including page content to the client terminal 1 (Step S116). The client terminal 1 performs display based on this response (step S117), and the user browses the content.

  After that, the first score calculation unit 307 of the search server 3 calculates the first score based on the query log of the query log DB 313 by batch processing or the like at a predetermined timing, and registers the calculated first score in the scoring DB 314 (step S121). The second score calculation unit 308 calculates a second score based on the query log in the query log DB 313, and registers the calculated second score in the scoring DB 314 (step S122). Details of the score calculation will be described later.

  Next, the navigational query extraction unit 309 extracts a navigational query based on the scoring result of the scoring DB 314 and registers it in the navigational query DB 312 (step S123). Details of the navigational query extraction will be described later.

  In this way, because the navigational query is extracted based on the query entered by the user and the site selection behavior from the search result, and reflected in the subsequent search, the possibility of fluctuation depending on the time It can respond appropriately to a certain navigational query.

  FIG. 5 is a diagram illustrating a processing example of calculation of the first score and the second score by the first score calculation unit 307 and the second score calculation unit 308.

  In FIG. 5, when the calculation process of the first score and the second score is started (step S201), the first score calculation unit 307 selects one query from the query log DB 313 (step S202).

  Next, the maximum number of clicks of the first to Nth URLs is set in the variable $ MAX (step S203), the average number of clicks is set in the variable $ AVE (step S204), and “next / The “re-search” click count is set in the variable $ BAD (step S205).

  Then, the first score is calculated by the following formula using the number of clicks of the first to Nth URLs except the one with the maximum number of clicks as the variable $ EACH in the calculation of Σ (step S206). The meaning of the formula will be described later.

First score = ($ MAX− $ AVE) ²
-Σ ($ EACH- $ AVE) ²
-($ BAD- $ AVE) ²
Next, the second score calculation unit 308 sets the maximum click frequency rate of the first to Nth URLs to the variable $ MAX _R (step S207), and sets the average click frequency rate to the variable $ AVE _R. set to (step S208), and set the "next / re-search" clicks rate to variable $ BAD _R (step S209).

Then, the second score is calculated by the following equation using the click frequency rate of the first to Nth URLs except the one with the maximum click frequency rate as the variable $ EACH _R in the calculation of Σ (step S210). The meaning of the formula will be described later.

Second score = ($ MAX _R − $ AVE _R ) ²
-Σ ($ EACH _R- $ AVE _R ) ²
_{- ($ BAD R - $ AVE} R) 2
Next, the URL of the query to be processed and the calculated first score, second score, and the maximum number of clicks (rate), the query of the scoring DB 314, the first score, the second score, the maximum number of clicks (rate) URL (Step S211).

  Next, it is determined whether or not the target query has been processed (step S212). If it has not been processed (No in step S212), the process returns to the selection of the next query (step S202), and has been processed ( In step S212, Yes) ends the calculation of the first score and the second score (step S213).

  FIG. 6 is a flowchart showing a processing example of navigational query extraction by the navigational query extraction unit 309.

  In FIG. 6, when the navigational query extraction process is started (step S301), the navigational query extraction unit 309 selects one query from the scoring DB 314 (step S302).

  Next, it is determined whether or not the first score is greater than or equal to a predetermined value (step S303). If it is greater than or equal to the predetermined value (Yes in step S303), it is subsequently determined whether or not the second score is greater than or equal to the predetermined value. (Step S304).

  If the second score is greater than or equal to a predetermined value (Yes in step S304), the URL of the query to be processed and the maximum number of clicks (rate) is registered in the navigational query and site URL of the navigational query DB 312 (step S305). .

  If the first score is not equal to or higher than the predetermined value (No in step S303), if the second score is not equal to or higher than the predetermined value (No in step S304), or after registration in the navigational query DB 312 (step S305), It is determined whether or not each query has been processed (step S306). If not processed (No in step S306), the process returns to the selection of the next query (step S302), and if processed (Yes in step S306). ) Ends the navigational query extraction process (step S307).

<The meaning of the first score and the second score>
Hereinafter, the meaning of the first score and the second score will be described.

  In the present invention, click log data in a search service is analyzed to automatically select a navigational query and a non-navigation query. At this time, in the above-described embodiment, the first score and the second score, which are indices indicating the probability of the navigational query, are calculated by applying the variance value calculation method (the first score and the second score are It is also a kind of variance value.), Those whose both exceed a predetermined threshold are extracted as navigational queries. Although the accuracy of extracting the navigational query is slightly lowered, only the first score may be used, and the first score exceeding a predetermined threshold may be extracted as the navigational query.

Navigational queries are
(1) One site in the search list is clicked in a concentrated manner.
(2) The target site is always included in the first page of the search list.
(3) “Next” and “Re-search” are not clicked.
It has the characteristic.

  Here, in the case of a navigational query, since there is a characteristic that “one site in the search list is clicked in a concentrated manner”, the determination of whether or not it is a navigational query has a variance value. The use is also considered suitable. The variance value is obtained by dividing the “sum of squared deviations from the average value” by the “number of elements”. 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. If all the elements are close to the average value, the variance value becomes small.

  However, if each element is disjoint, 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, in this embodiment, the variance value is not applied as it is, but the sum of deviation squares of other clicks is subtracted from the deviation square of the maximum clicks. As a result, in the case of a navigational query in which a certain site is clicked in a concentrated manner, the deviation squared value due to the number of clicks becomes larger, and the deviation squared sum of other clicks becomes relatively larger. It becomes smaller and the overall score value becomes larger. 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 the number of clicks is evenly clicked. As a result, it can be used as an index indicating a state where only one site is clicked in a concentrated manner.

  The number of clicks used for the calculation is displayed on the first page of the search list based on the characteristic of the navigational query that “the target site is always included in the first page of the search list” described above. The number can be limited to N (for example, 10).

In FIG. 7, the horizontal axis indicates the sites in the order of display order of the search list for a certain query, and the vertical axis indicates the number of selected clicks for each site. The first term “($ MAX− $ AVE) ² ” of the formula for calculating the first score corresponds to the first site in FIG. 7 (it is not always the first), and the second term “−” “Σ ($ EACH− $ AVE) ² ” is the total sum for the second to Nth sites in FIG.

On the other hand, the third term “− ($ BAD− $ AVE) ² ” in the calculation formula of the first score is based on the characteristic of the navigational query that “next” or “re-search” is not clicked ”. This is a reflection of the user's judgment that the search is not a navigational query due to the “re-search / re-search”.

  The second score is calculated based on the click frequency rate, whereas the first score is calculated based on the number of clicks (absolute number). In other words, since the score tends to be high in a query with a high search request, this is to remove the influence. 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) In the case of scoring only by the number of clicks, a query with a higher number of searches has a higher score.

  Therefore, scoring is performed with respect to the “shape of click distribution” 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 rate, the score of a query with a low impression will be high, so that it is not used alone, and a second score is predetermined for a case where the first score is larger 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.

<Modification>
Since the navigational query corresponds to being used instead of the bookmark, the click log corresponding to the query input by the user for the second time or later can be used for the calculation of the first score and the second score.

The flow of processing in this case is as follows.
(1) The search server 3 manages a search history for each user by logging in with a user ID.
(2) The click log collection unit 306 stores the click log in the query log DB 313 for each user ID.
(3) The first score calculation unit 307 and the second score calculation unit 308 calculate the first score and the second score without using the click log for the query input for the first time by the user for the calculation of the score.

  Thereby, the navigational query with respect to the site which the user browsed once can be extracted.

<Summary>
As described above, according to the embodiment of the present invention, a navigational query can be automatically extracted with high accuracy.

And by using the navigational query in the search service, for example,
(1) When the site corresponding to the navigational query is in the second and subsequent search results, the display order is changed and moved to the top to obtain a search result that meets the user's needs.
(2) The user's operability is improved by acquiring a query input at the client terminal, and suggesting and guiding a site corresponding to the query to the user.
It can be used for useful applications such as.

  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.

It is a figure which shows the structural example of the system concerning one Embodiment of this invention. It is a figure which shows the data structure example of each database. It is a sequence diagram which shows the process example of embodiment. It is a figure which shows the example of a screen display at the time of a search. It is a figure which shows the example of a process of calculation of a 1st score and a 2nd score. It is a flowchart which shows the process example of a navigational query extraction. It is explanatory drawing of the 1st term and the 2nd term in the calculation formula of the 1st score.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Client terminal 101 Query input part 102 Search execution instruction reception part 103 Query transmission part 104 Recommended site reception part 105 Search list reception part 106 Display control part 107 Site selection reception part 108 Redirect processing part 2 Network 3 Search server 301 Query reception part 302 Search processing unit 303 Recommended site transmission unit 304 Search list transmission unit 305 Redirect processing unit 306 Click log collection unit 307 First score calculation unit 308 Second score calculation unit 309 Navigational query extraction unit 311 Content DB
312 Navigational Query DB
313 Query log DB
314 Scoring DB
4 Web server

Claims (11)

Aggregating a click log indicating the history of individual selections for a plurality of sites from a search list for a query for each query;
Calculating a first variance based on the number of clicks per site for each query based on the click log;
A query extraction method comprising: a query in which the calculated first variance value is equal to or greater than a predetermined value; and a step of extracting site information having the highest number of clicks in a search list for the query. Aggregating a click log indicating the history of individual selections for a plurality of sites from a search list for a query for each query;
Calculating a first variance based on the number of clicks per site for each query based on the click log;
Calculating a second variance value based on a click rate for each site for each query based on the click log;
A query in which the calculated first variance value and the second variance value are each equal to or greater than a predetermined value, and a step of extracting site information having the highest number of clicks in a search list for the query. Feature query extraction method. In the query extraction method according to claim 1 or 2,
The step of calculating the first variance value includes calculating the first variance value by subtracting the sum of deviation squares of other clicks from the deviation square of the maximum number of clicks. . The query extraction method according to claim 3, wherein
On the screen displaying the search list for the query, a button for changing the search list is displayed.
The step of calculating the first variance value further includes calculating the first variance value by subtracting a deviation square of the number of clicks of the displayed button. The query extraction method according to claim 2,
The step of calculating the second variance value calculates the second variance value by subtracting the sum of deviation squares of other click frequency rates from the square of the deviation of the maximum click frequency rate. Extraction method. The query extraction method according to claim 5, wherein
A search list for the query and a button for changing the displayed search list are displayed on one screen.
The step of calculating the second variance value further includes calculating the second variance value by subtracting the deviation square of the click frequency rate of the displayed button. In the query extraction method according to any one of claims 1 to 6,
The step of counting the click log includes counting the click log for each user,
The step of calculating the variance value uses the click log corresponding to the query input by the user for the second time or later on the basis of the click log for each user to calculate the variance value. Means for aggregating, for each query, a click log indicating a history of individual selections for a plurality of sites from a search list for a query;
Means for calculating a first variance based on the number of clicks per site for each query based on the click log;
A query extraction apparatus comprising: a query in which the calculated first variance value is equal to or greater than a predetermined value; and means for extracting site information having the highest number of clicks in a search list for the query. Means for aggregating, for each query, a click log indicating a history of individual selections for a plurality of sites from a search list for a query;
Means for calculating a first variance based on the number of clicks per site for each query based on the click log;
Means for calculating a second variance based on the click rate for each site for each query based on the click log;
A query in which the calculated first variance value and the second variance value are each equal to or greater than a predetermined value, and means for extracting site information having the highest number of clicks in a search list for the query. A feature query extraction device. A computer constituting the processing device,
Means for aggregating, for each query, a click log indicating a history of individual selections for a plurality of sites from a search list for a query;
Means for calculating a first variance based on the number of clicks per site for each query based on the click log;
Means for extracting a query in which the calculated first variance value is a predetermined value or more, and site information having the highest number of clicks in a search list for the query;
Query extraction program to function as. A computer constituting the processing device,
Means for aggregating, for each query, a click log indicating a history of individual selections for a plurality of sites from a search list for a query;
Means for calculating a first variance based on the number of clicks per site for each query based on the click log;
Means for calculating a second variance based on a click rate for each site for each query based on the click log;
Means for extracting a query in which the calculated first variance value and the second variance value are each equal to or greater than a predetermined value, and site information having the highest number of clicks in a search list for the query;
Query extraction program to function as.

Images