JP2011154466A - Retrieval result ranking method and system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To perform retrieval result ranking with the tastes of users reflected thereon. <P>SOLUTION: A retrieval result ranking method includes: a score calculation step for calculating scores to be used for the ranking in order of displaying in a retrieval result through the use of evaluation values tabulated by user retrieval log information stored in a retrieving device 20; and a retrieval result display step for performing the ranking by using the scores and displaying the retrieval result in order of the ranks. The method further includes: a user-specific revisit number of times tabulation step (step S116) for tabulating the number of times of user-specific revisits being the times of the revisits of the users to the retrieval result, based on the times of appearance of the same user identifiers and tabulation keys, on the basis of the retrieval log information in order to tabulate the evaluation values; and an evaluation value tabulation step (step S118) for tabulating the evaluation values, based on the number of user-specific revisits which is tabulated in the user-specific revisit number tabulation step. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a search result ranking method and a search result ranking system.

  Ranking (ranking) of the display order of URLs displayed as search results on the search web page is scored using various indexes. A well-known index is one that uses the characteristics of the page itself, such as the number of times a search word appears in the page or a score calculated from the link relationship between pages in the web space. As another index, a user's click log (hereinafter also referred to as “search click log” or “search log information”) for the search result output by the search engine can be used. This index is useful as user feedback information for search results, but it is affected by factors that are not directly related to the suitability of search results, such as the order of the results, so it is useful to indicate the suitability of search results from the search click log. It is important to extract appropriate indicators appropriately.

  Patent Document 1 discloses a technique for improving a search result using a search click log. As an index that can be extracted from the search click log, the click-through rate of the search page (ratio of the number of times the result is clicked with respect to the number of times the search result is displayed) is described.

US Pat. No. 6,493,702

  However, the search results are not necessarily clicked after the relevance of the search has been thoroughly investigated by the user, and the click-through rate is appropriate for the search query because it is more likely to be clicked as it is displayed at the top. It may not be enough as an index to express

  The present invention has been made to solve such a problem, and an index indicating suitability for a search query is extracted from a search click log, and search result ranking that can appropriately rank pages is performed. It is an object to provide a method and a search result ranking system.

  In order to solve the above-mentioned problems, the search result ranking method of the present invention uses the evaluation values accumulated from the user's search log information accumulated in the search device, and displays the search results displayed on the search result display screen. A score calculation step for calculating scores used for ranking the display order, and a search result display step for performing the ranking using the scores and displaying the search results on the search result display screen in the rank order For each of the users, the number of times the user has revisited the search result based on the number of times the same user identifier and count key have appeared based on the search log information. An evaluation value for totaling the evaluation values based on the number of revisiting for each user for counting the number of revisits and the number of revisiting for each user calculated in the number of revisiting for each user Meter steps, and further comprising a.

  The search result ranking system according to the present invention includes a search log information storage unit that stores user search log information, and an evaluation value total that totals evaluation values from the search log information stored in the search log information storage unit. Module, score calculation means for calculating a score used for ranking the display order of the search results displayed on the search result display screen using the evaluation values totaled by the evaluation value totaling module, and using the scores Search result display means for performing ranking and displaying the search results in the order of ranking on the search result display screen, and the evaluation value aggregation module is based on the search log information. Based on the number of appearances of the user identifier and the total key, the total number of visits by user totaling the number of visits by user, which is the number of times the user has revisited the search result. If, based on the number of user-specific revisit that the user-specific revisits tally means and aggregate, characterized in that it comprises an evaluation value collecting unit that aggregates the evaluation value.

  The search result ranking method of the present invention uses the evaluation values aggregated from the user's search log information accumulated in the search device, and ranks the display order of the search results displayed on the search result display screen. A score calculation step for calculating a score used in the above, a search result display step for performing the ranking using the score and displaying the search results in the order of ranking on the search result display screen, Based on the search log information filtering step for filtering the search log information and the filtered search log information in order to total the evaluation values, based on the number of times the same user identifier and total key have appeared, the user Summing the number of visits by user, which is the number of times the user has revisited the search result, and the number of visits by user Based on the number of user-specific revisit that are aggregated in meter steps, and further comprising and a summary data step of aggregating the evaluation value.

  The search result ranking system according to the present invention includes a search log information storage unit that stores user search log information, and an evaluation value total that totals evaluation values from the search log information stored in the search log information storage unit. Module, score calculation means for calculating a score used for ranking the display order of the search results displayed on the search result display screen using the evaluation values totaled by the evaluation value totaling module, and using the scores Search result display means for performing ranking and displaying the search results on the search result display screen in the order of the ranking, and the evaluation value aggregation module includes search log information for filtering the search log information Based on the number of appearances of the same user identifier and total key based on the filtering means and the filtered search log information. Based on the number of revisits for each user, which is the number of times the user has revisited the search results, and the number of revisits for each user counted by the number of revisits for each user, the evaluation Evaluation value totaling means for totaling the values.

  According to the search result ranking method and search result ranking system of the present invention, the number of times the user revisited the search results based on the number of times the same user identifier and total key appeared based on the search log information. By counting the number of revisits for each user, it is possible to specify the number of revisits to a page that the user is satisfied with or at least cares from the search results. By calculating the score of each page using the number of visits by user and the evaluation value collected by this method, and using the score for ranking the search results, it is possible to rank the search results more reflecting the user's preference. It becomes.

  Further, in the present invention, the search log information filtering step includes a search session specifying step for specifying a search session representing a series of search behaviors of the user. In the search session specifying step, the search session is included in the search result. It is preferable that the search session is specified so as to include a certain number or less of records indicating access.

  According to this invention, a series of search actions of a user can be specified by specifying a search session. In addition, by limiting the number of records included in the search session to a certain number or less, exceptional search behavior that accesses many search results can be excluded.

  Further, in the present invention, the search log information filtering step includes a search session specifying step for specifying a search session representing a series of search behaviors of the user, and in the search session specifying step, a record in which the total key is duplicated is included. Preferably, the search session is identified so as to be excluded.

  According to this invention, a series of search actions of a user can be specified by specifying a search session. Also, by excluding duplicate data, the amount of subsequent processing can be reduced and the processing efficiency can be increased.

  Further, in the present invention, the search log information filtering step includes a search session specifying step for specifying a search session representing a series of search behaviors of the user. In the search session specifying step, the search session is included in the search result. It is preferable that the search session is specified so as to include a certain number or less of records indicating that access has been made, and to exclude records in which the aggregation keys are duplicated.

  According to this invention, a series of search actions of a user can be specified by specifying a search session. In addition, by limiting the number of records included in the search session to a certain number or less, exceptional search behavior that accesses many search results can be excluded. Also, by excluding duplicate data, the amount of subsequent processing can be reduced and the processing efficiency can be increased.

  In the present invention, the search log information filtering step filters the search session so that a predetermined time interval exists between the search sessions after the search session specifying process by the search session specifying step. Preferably it further comprises a search session filtering step.

  According to the present invention, a series of search actions continuously performed within a predetermined time interval can be collectively specified as one search action (search session). Note that “between each search session” means, for example, between the access time of the last record of the previous search session and the access time of the first record of the subsequent search session.

  Further, in the present invention, the search session is a series of the records in the search behavior by the same terminal, the search request record being a record indicating the search request in the search behavior as a starting point, and the search that appears next It is preferable that the last record in the entire search log information is set as the end point.

  According to the present invention, a series of search behaviors of a user starting from a search request can be specified as one search behavior. Specifically, it is possible to identify a series of actions that have browsed a search result related to the search term returned from the search device, starting with the user entering a search term into the search device.

  In the present invention, it is preferable that the search log information filtering step further includes a first record filtering step of extracting the records belonging to a specific genre.

  According to the present invention, by extracting only records belonging to a specific genre, it is possible to suppress the calculation amount of the subsequent processing and increase the processing efficiency. In addition, the evaluation value according to the present invention can be given only to the genre effective in ranking the search results.

  Further, in the present invention, the search log information filtering step further includes a second record filtering step of filtering the records so that a time interval between the records having the same user identifier and the total key is equal to or greater than a predetermined value. It is preferable to provide.

  According to this invention, it is possible to improve the reliability of the user-by-user revisit count counting process by setting a fixed time interval for the user revisit.

  In the present invention, it is preferable that the total key is one or both of a search word and attribute information of the search result.

  According to the present invention, a specific example of information used as a total key is provided.

  Further, in the present invention, in the evaluation value totaling step, the search word and the attribute information of the search result are used as a key for the number of revisits for each user counted in the total number of revisits for each user. It is preferable to tabulate evaluation values.

  According to the present invention, the score of each page is calculated using the search word and the attribute information of the search result as a key for the number of revisits for each user, and the score is ranked in the URL of the search result. By using the search results, it is possible to rank search results more reflecting the user's preference.

  Further, in the present invention, in the evaluation value totaling step, the evaluation values are totaled using the attribute information of the search result as a key for the number of revisiting by user totaled in the number of revisiting by user. It is preferable to do.

  According to the present invention, the score of each page is calculated using the attribute information of the search result as a key for the number of visits by user, and the score is used for ranking the URL of the search result. Thus, it is possible to rank search results more reflecting the user's preference.

  In the present invention, it is preferable that in the evaluation value totaling step, an average value of the number of revisits for each user for the key is totaled as the evaluation value.

  According to this invention, the average number of visits to the search result is used as an evaluation value, the score of each page is calculated, and the score is used for ranking the URL of the search result, thereby reflecting the user's preference more. The search result ranking can be performed.

  In the present invention, it is preferable that in the evaluation value totaling step, the number of users revisited to the search result is totaled as the evaluation value.

  According to this invention, the number of users revisited to the search result is used as an evaluation value, the score of each page is calculated, and the score is used for ranking the URL of the search result, thereby further enhancing the user preference. The reflected search result ranking can be performed.

  In the present invention, it is preferable that, in the evaluation value aggregation step, the aggregation is performed when the number of visits by user is a predetermined value or more.

  According to this invention, the reliability of the aggregated evaluation values can be increased.

  In the present invention, it is preferable that in the evaluation value totaling step, a value obtained by dividing the totaled evaluation value by the number of unique users for the key is totaled as a final evaluation value.

  According to the present invention, the evaluation value can be acquired as a normalized value of 0 to 1.

  According to the present invention, it is possible to provide a search result ranking method and a search result ranking system capable of ranking search results reflecting user preferences.

1 is a schematic configuration diagram of a search result ranking system 1 according to an embodiment of the present invention. 2 is a hardware configuration diagram of an evaluation value totaling device 10 and a search device 20 that constitute a search result ranking system 1. FIG. It is a sequence diagram for demonstrating the whole flow of the control processing of the search result ranking system 1. FIG. It is a figure which shows the example of the search log information accumulate | stored in FIG.3 S130. It is a figure which shows the example of the search result output in step S150 of FIG. It is a figure which shows the example of the search log information accumulate | stored in step S170 of FIG. It is a figure which shows the example of the search log information accumulate | stored in the search log information storage part 26 through the procedure of step S100-step S180 of FIG. It is a flowchart which shows the evaluation value total process of step S110 of FIG. 3 in detail. It is a figure which shows the example of search session ID provided with respect to the search log information of FIG. 7 according to the procedure of the evaluation value totaling process shown in FIG. It is a figure which shows the example of the revisit record extracted in step S116 of FIG. It is a figure which shows the number of revisit clicks as an example of the evaluation value totaled by step S118 of FIG. It is a figure which shows the average number of revisit clicks and the number of revisited unic users as an example of the evaluation value totaled by step S118 of FIG.

  Hereinafter, preferred embodiments of a search result ranking method and a search result ranking system according to the present invention will be described in detail with reference to the accompanying drawings. In the description of the drawings, the same elements are denoted by the same reference numerals, and redundant description is omitted.

(Configuration of search result ranking system 1)
FIG. 1 is a diagram showing a configuration of a search result ranking system 1 according to an embodiment of the present invention. In response to a request from the client terminal 30, the search result ranking system 1 ranks the display order of URLs displayed as search results on the search web page, and reflects the search results on the search web page. It has a function of outputting to the client terminal 30 and displaying it. The client terminal 30 can make a content request to a communication network 40 such as WWW (World Wide Web) using a protocol such as HTTP and browse the content. As shown in FIG. 1, the search result ranking system 1 includes an evaluation value totaling device 10 (corresponding to an “evaluation value totaling module” in the claims) and a search device 20, and includes a client terminal 30. Are communicably connected through the communication network 40. In FIG. 1, the evaluation value totaling device 10 and the search device 20 are illustrated as separate devices, but the present invention is not limited to this, and the evaluation value totaling device 10 and the search device 20 are one device (not shown). It may be configured as

  FIG. 2 is a hardware configuration diagram of the evaluation value totaling device 10 and the search device 20 (or the device in the case of a single device) constituting the search result ranking system 1. As shown in FIG. 2, the evaluation value totaling device 10 and the search device 20 are physically composed of a main storage device such as a CPU, a ROM and a RAM, an input device such as a keyboard and a mouse, an output device such as a display, and a client. It is configured as a normal computer system including a communication module such as a network card for transmitting and receiving data to and from other devices such as the terminal 30 and an auxiliary storage device such as a hard disk. Each function of the evaluation value totaling device 10 and the search device 20 to be described later reads predetermined computer software on hardware such as a CPU, a ROM, a RAM, etc., so that an input device, an output device, This is realized by operating the communication module and reading and writing data in the main storage device and the auxiliary storage device.

  Returning to FIG. 1, the evaluation value totaling apparatus 10 totals the evaluation values from the search log information stored in the search log information storage unit 26 (corresponding to “search log information storage unit” in the claims) of the search apparatus 20. To do. The evaluation value in the present embodiment is used when calculating a score for displaying the search results, and includes the number of clicks to be revisited later. As shown in FIG. 1, the evaluation value totaling apparatus 10 includes a search log information filtering unit 11 (corresponding to “search log information filtering means” in the claims), a revisit click totalizing unit 12 (“users in the claims” And a totaling result storage unit 13).

  The search log information filtering unit 11 filters the search log information stored in the search log information storage unit 26 and includes a search session specifying unit 111. The search session specifying unit 111 represents a series of search behaviors of the user, and is included in a plurality of search log information (see, for example, FIG. 7 described later) stored in the search log information storage unit 26 of the search device 20. It has a function to specify the search session. The plurality of search sessions include a series of records from the search request to the end of the search action, each accessed by the same client terminal 30. Specifically, after the search log information is sorted in the order of users and time, a search session can be specified by extracting a search request or search result click by the same user and the same search query. Details of filtering processing by the search log information filtering unit 11 and details of a search session specifying procedure by the search session specifying unit 111 will be described later.

  Based on the search log information filtered by the search log information filtering unit 11, the revisit click counting unit 12 has the same user identifier and total key (for example, one or both of the search word and the attribute information of the search result). Based on the number of appearances, the number of revisits for each user, which is the number of times the user has revisited the search result in relation to the search term, is totalized, and the evaluation values are totalized based on the total number of revisits for each user. Specifically, the revisit click counting unit 12 includes a revisit record (hereinafter, referred to as a record indicating that the user has accessed the search result a plurality of times, among the records constituting each search session specified by the search session specifying unit 111). (Also referred to as “revisit click”), and the evaluation values are tabulated based on the extracted revisit record. In other words, the revisit click counting unit 12 extracts only a plurality of search result clicks (revisit clicks) from each search session, and then totals the number of clicks and the number of unique users for each URL as evaluation values. The types of evaluation values to be summarized will be described later.

  The tabulation result storage unit 13 has a function of storing tabulated data from the revisit click tabulation unit 12 in the score calculation data storage unit 27.

  In response to a search query requested from the client terminal 30, the search device 20 returns a link set to a page matching it as a result. The search device 20 includes a transfer unit 21, a search result output unit 22 (corresponding to “search result display means” in the claims), a search reception unit 23, a search log information recording unit 24, and a score calculation unit 25 (claims). A search log information storage unit 26 (corresponding to “search log information storage unit” in the claims), and a score calculation data storage unit 27.

  The search reception unit 23 has a function of receiving a search query requested from the client terminal 30 and transferring the search query to the score calculation unit 25, and a function of transferring acceptance information from the client terminal 30 to the search log information recording unit 24. Have.

  The score calculation data storage unit 27 has a function of storing score calculation data (evaluation value) used for determining the order of search and a function of outputting an evaluation value to the score calculation unit 25.

  The score calculation unit 25 uses the evaluation value stored in the score calculation data storage unit 27 in order to determine content that matches the search query, and displays a search result display screen (not shown) of the client terminal 30. ) Has a function of calculating a score used for ranking the display order of the search results displayed. Usually, the score calculation unit 25 scores the degree of fitness of the document with respect to the search query using a plurality of evaluation indexes. The number of revisited clicks according to the present invention is one of evaluation indexes. An example of simple score calculation that can be performed by the score calculation unit 25 is to calculate the number of revisit clicks on a document including a search query as a score. The score calculation unit 25 has a function of outputting the calculated score to the search result output unit 22.

  The search result output unit 22 ranks the search results using the scores, and displays the search results on the search result display screen of the client terminal 30 in the order of the ranks. That is, when the search result output unit 22 outputs a link set to the content (URL) to the client terminal 30 as a search result, the search result output unit 22 selects the content in descending order based on the score calculated by the score calculation unit 25. It has the function to output side by side. In the search result, the URL of the result page is not described as a direct link, but is described as a parameter of an access URL to the transfer unit 21. When the user clicks on the result, the client terminal 30 once accesses the transfer unit 21. Next, the transfer unit 21 refers to the parameter included in the access URL and requests the client terminal 30 to transfer the result page. Finally, the client terminal 30 accesses the result page.

  When there is an access from the client terminal 30, the transfer unit 21 refers to the parameter included in the access URL and requests the client terminal 30 to transfer to the result page, and clicks the search result in the search log information recording unit 24. It has a function of outputting information (information indicating that the user has accessed the URL).

  The search log information recording unit 24 has a function of recording the search query from the search receiving unit 23 and the search result click information from the transfer unit 21 in the search log information storage unit 26 as user search log information.

  The search log information storage unit 26 stores search queries by the client terminal 30 and click information on search results as user search log information. By analyzing the search log information stored in the search log information storage unit 26 by the evaluation value totaling device 10, the number of revisited clicks in the present invention is totaled.

(Operation of search result ranking system 1)
The search result ranking method performed by the search result ranking system 1 according to the embodiment of the present invention will be described with reference to FIGS.

(Overall operation)
First, the overall flow of operations by the search result ranking system 1 will be described. FIG. 3 is a sequence diagram for explaining the overall flow of control processing of the search result ranking system 1 according to the embodiment of the present invention.

  In the search result ranking system 1, as preprocessing, the evaluation value totaling device 10 acquires user search log information collected and stored in advance by the search log information storage unit 26 of the search device 20 (step S100). After the evaluation values are aggregated from the acquired search log information, the evaluation values after the aggregation are stored in the score calculation data storage unit 27 of the search device 20 (step S110). The details of the evaluation value aggregation method will be described in detail after describing the search log information collection / accumulation method shown in steps S120 to S180 below.

  The client terminal 30 makes a search request to the search device 20 (step S120). The search reception unit 23 of the search device 20 transfers this search request to the search log information recording unit 24, and the search log information recording unit 24 stores it as search log information in the search log information storage unit 26 (step S130). An example of the accumulated data is shown in FIG. In this example, there are items of user ID, time stamp, type, search term, and transition destination URL. The user ID identifies a different user by some method such as a browser cookie, a user ID notified to the search device 20 by an Internet provider, or an ID issued by the search device 20 when the search device 20 is accessed. is there. The time stamp is the date and time when the client terminal 30 accessed the search device 20, the type is the action type by the client terminal 30, the search word is the word input by the user and requested to be searched, and the transition destination URL is the search The URL of the page accessed by clicking the search result output by the device 20 (here, it is blank because there was only a search request) is shown.

  Next, returning to FIG. 3, the score calculation unit 25 of the search device 20 calculates the degree of matching between the search term requested in step S120 and the page as a score (step S140, “score calculation step in claims”). ”). As described above, the score is used for ranking the display order of search results displayed on the search result display screen of the client terminal 30. In the score calculation, the number of revisited clicks in the present invention is used (other indexes can also be used).

  Next, the search result output unit 22 ranks the search results using the score calculated in step S140, outputs the search results to the client terminal 30, and displays the search result display screen of the client terminal 30 on the search result display screen. The search results are displayed in order of rank (step S150, corresponding to “search result display step” in the claims). An example of the search result is shown in FIG. Each result (“aa”, “aaa”, etc.) is a link clickable by the user. The link destination address indicates the transfer unit 21 of the search device 20, and the search term and the URL of the result page are included as parameters. For example, the route address of the transfer unit 21 is “www.docomo.co.jp”, the search term is “AAA”, and the result page address is “http://www.AAA.ne.jp/aaa.htm”. Then, “http://www.docomo.co.jp/index.html?word=AAA&result=http://www.AAA.ne.jp/aaa.htm” is represented. Thus, the search term and the transition destination address are URL encoded.

  Next, returning to FIG. 3, when the user clicks the search result, the client terminal 30 accesses the transfer unit 21 of the search device 20, and is URL-encoded as the above parameter together with information that the user clicked the search result. Is transmitted from the transfer unit 21 to the search log information recording unit 24 (step S160).

  Next, the search log information recording unit 24 collects and accumulates this click information as search log information (step S170). FIG. 6 shows an example of search log information at this point.

  At the same time, the transfer unit 21 requests the client terminal 30 to access the result page (Step S180). The client terminal 30 accesses the result page according to the instruction. After accessing the result page, the user can return to the search result and browse other results.

  A series of search log information ranging from the search request to the browsing of the search result is left in the search log information storage unit 26 by the procedure from step S100 to step S180 described above. An example of the collected search log information is shown in FIG.

(Details of evaluation value aggregation processing)
Subsequently, the evaluation value totaling process (step S110) for analyzing the search log information stored in the search log information storage unit 26 and totaling the evaluation values including the number of revisited clicks will be described in detail. FIG. 8 is a flowchart of the evaluation value totaling process.

  As shown in FIG. 8, first, the search log information filtering unit 11 of the evaluation value totaling apparatus 10 acquires search log information from the search log information storage unit 26 (step S112).

  Next, the search log information filtering unit 11 filters the search log information acquired in step S112 (step S114, corresponding to the “search log information filtering step” in the claims).

Specifically, the search session specifying unit 111 belonging to the search log information filtering unit 11 specifies a search session from the search log information acquired in step S112 (corresponding to “search session specifying step” in the claims). The procedure for specifying a search session is shown below.
Procedure 1: Sort search log information by user ID and time stamp.
Procedure 2: A record having type = “search request” is specified as a start point, and records up to an end point among records having the same user ID as the start point are specified as the same search session ID. As an end point, it is determined whether the next record that appears next = record of “search request” or the last record is the end point.

  In other words, a search session is a series of records in a search action by the same terminal, starting from a search request record that is a record indicating a search request, and is a record immediately preceding the search request record that appears next, or a search The last record in the entire log information is the end point. FIG. 9 shows an example of the search session ID assigned to the search log information of FIG. 7 according to the above procedure. In FIG. 9, the search request record is a record of type = “search request”, and is the first, fourth, etc. record from the top.

  Next, based on the search session specified in step S114, the same user identifier (eg, “user ID” in the example of FIG. 9), search word (eg, “search word” in the example of FIG. 9). ), And the number of revisits for each user, which is the number of times the user revisited the search results in relation to the search word, based on the number of times the search result attribute information (for example, “transition destination URL” in FIG. 9) appears (Step S116, equivalent to “revisiting count for each user” in the claims). Specifically, the revisit click counting unit 12 extracts a revisit record that is a record indicating that the user has accessed the search result a plurality of times from among the records constituting each search session specified by the search session specifying unit 111. Based on the extracted revisit record, the number of revisits by user is tabulated.

  Finally, the evaluation values are totaled based on the number of visits by user totaled in step S116 (step S118, corresponding to the “evaluation value totaling step” in the claims).

  As a specific example of the processing described above, FIG. 10 shows an example of the revisit record extracted by the revisit click counting unit 12 in step S116. In this example, the user with the user ID = 1000 visits the page “http://www.AAA.ne.jp/aaa.htm” for two search sessions with the search term “AAA”. Therefore, the number of revisited clicks is 1. This is a value obtained by subtracting 1 (first access) from the total number of accesses. Similarly, the user with user ID = 1001 has visited “AAA” on the page “http://www.AAA.ne.jp/aaa.htm” over three search sessions, and the number of revisited clicks is 2. Become. FIG. 11 shows the result of counting the number of revisited clicks.

  In step S118, the evaluation values are totaled using the search term and / or attribute information of the search result as a key for the number of visits by user totaled in step S116. As one of the evaluation values tabulated in step S118, there is an average number of revisit clicks (corresponding to “an average value of the number of revisits for each user” in the claims). An example of the total number of average revisit clicks is shown in FIG. In the example of FIG. 12, the number of revisited clicks on the page “http://www.AAA.ne.jp/aaa.htm” for the search term “AAA” in FIG. Is 1.5.

  Further, as one of the operations performed in step S118, the number of revisited unic users (number of revisited UUs) is counted as one of the evaluation values. The number of unvisited unic users is the number of users who have revisited the search results. The number of unvisited unic users is tabulated for each search term and search result URL. In FIG. 10, the users who revisited the page “http://www.AAA.ne.jp/aaa.htm” with respect to the search term “AAA” are the user with ID = 1000 and the user with ID = 1001. Therefore, the number of revising unic users is two. FIG. 12 shows the result of counting the number of revisited unic users.

  Moreover, in step S118, it is good also as a number which divided the number of unique users revisited by the number of unique users by a total key as an evaluation value. Thereby, an evaluation value is acquirable as a normalized value of 0-1. For example, in the example of FIG. 10, the number of unique users using the search term and the transition destination URL as keys is two. Therefore, 1.0 is obtained as a value obtained by dividing the number of revisiting unique users by the number of unique users.

  By calculating the score of each page using the evaluation values tabulated by the above method and using the score for ranking the search results, it is possible to rank the search results more reflecting the user's preference.

  The evaluation values tabulated as described above are stored in the score calculation data storage unit 27 by the tabulation result storage unit 13. In step S118, the aggregation process may be performed only when the number of revisited clicks on the search result is equal to or greater than a certain value. Thereby, the reliability of the aggregated evaluation values can be increased. For example, if the number of revisited clicks is 2 or more, the number of revisited clicks for the user ID = 1000 is 1 in the example of FIG. Therefore, the average number of revisited clicks is 2, the number of revisited unic users is 1, and the tabulation results are different from those illustrated in FIG.

  Next, operations and effects of the search result ranking system 1 according to the present embodiment will be described. According to the search result ranking system 1 of the present embodiment, based on the search log information, the user can search for a search word based on the number of times the same user identifier and total key (for example, search word and search result attribute information) appear. By counting the number of visits by user, which is the number of times the search results have been revisited in relation to, the number of visits to the page that the user has satisfied or at least interested in the search results is identified. be able to. By calculating the score of each page using the number of visits by user and the evaluation value collected by this method, and using the score for ranking the search results, it is possible to rank the search results more reflecting the user's preference. It becomes.

  In addition, it is possible to search with the same search word over a plurality of search sessions and extract a pattern that transitions to the same page. As a result, it is possible to specify a page having a high degree of matching with respect to the search word, express the degree of matching with the total value, and use it for the search ranking.

  In addition, a series of search actions of a user starting from a search request can be specified as one search action. Specifically, it is possible to specify a series of actions that have browsed the search result related to the search term returned from the search device 20, starting from the user entering a search term into the search device 20.

(Modification)
As mentioned above, although preferred embodiment of this invention was described, it cannot be overemphasized that this invention is not limited to the said embodiment. Other aspects of this embodiment will be exemplified below.

(Modification, part 1)
For example, in step S114 for specifying a search session, the search session may be specified so as to include a certain number or less of records indicating that the search session has accessed the search result. For example, it can be said that the example of FIG. 9 specifies the search session so that the search session includes two or less “result click” records. In this case, by limiting the number of records included in the search session to a certain number or less, exceptional search behavior that accesses many search results can be excluded.

(Modification, part 2)
Further, for example, in step S114 for specifying a search session, the search session may be specified so as to exclude a record having a duplicate key such as attribute information of a search result. In this case, by excluding duplicate data, it is possible to reduce the calculation amount of the subsequent processing and increase the processing efficiency. In this case, the “search result attribute information” includes, for example, a part of the structure of the transition destination URL including the above-described transition destination URL (search result URL) (for example, the URL host part or URL). And the category to which the transition destination URL belongs. As a specific example of the category, for example, in the case of i-mode (registered trademark), a menu ID to which the URL belongs can be cited. In addition, a specific genre (search portal directory) classified manually or a specific genre assigned by genre determination software is included.

(Modification, part 3)
In addition, for example, in step S114 for specifying a search session, the search session may be specified so as to have the two modified examples. In other words, in step S114, the search session is set so as to include a certain number or less of records indicating that the search session has accessed the search results, and to exclude records with overlapping counting keys such as attribute information of the search results. You may specify. In this case, by limiting the number of records included in the search session to a certain number or less, exceptional search behavior accessing many search results can be excluded, and by eliminating duplicate data, It is possible to reduce the calculation amount of the subsequent processing and increase the processing efficiency.

(Modification, part 4)
Further, for example, the step S114 of filtering the search log information includes a step of filtering the search session so that a predetermined time interval exists between the search sessions after the search session specifying process (see “ (Corresponding to “search session filtering step”). In this case, a series of search actions continuously performed within a predetermined time interval can be collectively specified as one search action (search session). Note that “between each search session” means, for example, between the access time of the last record of the previous search session and the access time of the first record of the subsequent search session. For example, in the example of FIG. 9, if each search session is 10 minutes, for example, there is only a time interval of about 6 minutes between the ID2 session and the ID3 session. 9 will be excluded.

(Modification, part 5)
For example, step S114 for filtering search log information may further include a step of extracting records belonging to a specific genre (corresponding to “first record filtering step” in the claims). In this case, by extracting only the records belonging to a specific genre, the calculation amount of the subsequent processing can be suppressed and the processing efficiency can be improved. In addition, the evaluation value according to the present invention can be given only to the genre effective in ranking the search results. The “specific genre” mentioned here includes, for example, the type of search source. Search results often include search results from a plurality of information sources such as images, news, and blogs in addition to normal results. When it is desired to remove a genre that is not correlated with the click order or the number of clicks, or when it is desired to obtain only a genre whose evaluation value from the click log is particularly useful, an evaluation value only for a specific genre is calculated. However, since all click logs are necessary for session identification, after the session is identified, revisited clicks belonging to the corresponding genre are extracted.

(Modification, part 6)
Further, for example, the step S114 for filtering the search log information is a step of filtering the records so that the time interval between records having the same user identifier and the total key is equal to or greater than a certain value (“second in the claims” (Corresponding to “record filtering step”). In this case, by setting a certain time interval for the user's revisit, it is possible to improve the reliability of the revisiting count processing for each user.

  Specifically, the records to be counted are those for which a certain time or more has passed since the access time (time stamp) of the record having the same user ID, search word, and transition destination URL before that record. To filter the records. For example, if it is a condition to be counted that 20 minutes have passed, in the example of FIG. 10, in the record of user ID = 1000, the search session ID = 1, the search term “AAA”, From the record of the page “http://www.AAA.ne.jp/aaa.htm”, the search session ID = 2, the search term “AAA”, and the page “http://www.AAA.ne.jp/aaa. Since the time interval to the “htm” record is 20 minutes or less, the latter record is not subject to aggregation.

(Modification, Part 7)
Moreover, the search session specifying method by the search session specifying unit 111 of the evaluation value totalizing apparatus 10 is not limited to the above. For example, the search session can be specified by the following procedure.
Procedure 1: Sort search log information by user ID and time stamp.
Procedure 2: A record having type = “search request” is specified as a start point, and records up to an end point among records having the same user ID as the start point are specified as the same search session ID. The end point is a record of type = “search request” that appears after the start point, and the search word is not a partial character string of the search word of the start point, or the search word of the start point is not a partial character string of the search word The last record is judged as the end point or the last record.

  That is, the search session in the modified example 7 is a series of records in a search action by the same terminal, starting from a search request record that is a record indicating a search request, and a search word and a part input in the search request at the start point The end point is the record immediately before another search request record by a search term not related to the character string or the last record in the entire search log information.

(Modification, Part 8)
Further, the search session specifying unit 111 of the evaluation value totaling apparatus 10 can specify the search session according to the following procedure.
Procedure 1: Sort search log information by user ID and time stamp.
Procedure 2: Among records having the same user ID, a series of records sequentially recorded with a time interval equal to or less than a predetermined time is set as the same search session.

(Variation, part 9)
Further, the search session specifying unit 111 of the evaluation value totaling apparatus 10 can specify the search session according to the following procedure.
Procedure 1: Sort search log information by user ID and time stamp.
Procedure 2: Among records having the same user ID, a series of records sequentially recorded with a time interval equal to or less than a predetermined time is set as the same search session. However, it is determined that the record having the type = “search request” is set as the start point and the next record that appears next == “search request” or the last record is set as the end point.

  That is, the search session in the modified example 9 is a series of records in search behavior by the same terminal, each of which is a series of records sequentially recorded with a time interval equal to or less than a predetermined time, and a record indicating a search request. The starting point is one of the records indicating that the search result has been accessed (one record before the time interval equal to or longer than the predetermined time or the last record in the entire search log information).

(Modification, Part 10)
Further, the search session specifying unit 111 of the evaluation value totaling apparatus 10 can specify the search session according to the following procedure.
Procedure 1: Sort search log information by user ID and time stamp.
Procedure 2: A record having type = “search request” is specified as a start point, and records up to an end point among records having the same user ID as the start point are specified as the same search session ID. As an end point, it is determined whether the next record that appears next = record of “search request” or the last record is the end point.
Procedure 3: A search request record is excluded from the search sessions sequentially specified in Procedure 2.
Step 4: When the time interval between the plurality of search sessions after the search request record is excluded in Step 3 is equal to or less than a predetermined time, the plurality of search sessions are integrated as one search session.

DESCRIPTION OF SYMBOLS 1 ... Search result ranking system, 10 ... Evaluation value totaling apparatus, 111 ... Search session specific | specification part, 12 ... Revisit click totalization part, 13 ... Total result storage part, 20 ... Search apparatus, 21 ... Transfer part, 22 ... Search result Output unit, 23 ... Search reception unit, 24 ... Search log information recording unit, 25 ... Score calculation unit, 26 ... Search log information storage unit, 27 ... Score calculation data storage unit, 30 ... Client terminal, 40 ... Communication network.

Claims (18)

A score calculation step for calculating a score used for ranking the display order of the search results displayed on the search result display screen using the evaluation values aggregated from the search log information of the user accumulated in the search device;
A search result display step for performing ranking using the score and displaying the search results in the order of ranking on the search result display screen;
With
In order to aggregate the evaluation values,
Based on the search log information, based on the number of times that the same user identifier and total key have appeared, the number of revisit by user totaling the number of visits by user, which is the number of times the user has revisited the search results,
Based on the number of revisits for each user counted in the number of revisits for each user, an evaluation value totaling step for totaling the evaluation values;
A search result ranking method, further comprising: A score calculation step for calculating a score used for ranking the display order of the search results displayed on the search result display screen using the evaluation values aggregated from the search log information of the user accumulated in the search device;
A search result display step for performing ranking using the score and displaying the search results in the order of ranking on the search result display screen;
With
In order to aggregate the evaluation values,
A search log information filtering step for filtering the search log information;
Based on the filtered search log information, based on the number of appearances of the same user identifier and total key, the total number of revisits by user that totals the number of revisits by user, which is the number of times the user has revisited the search results. Steps,
Based on the number of revisits for each user counted in the number of revisits for each user, an evaluation value totaling step for totaling the evaluation values;
A search result ranking method, further comprising: The search log information filtering step includes a search session specifying step for specifying a search session representing a series of search behaviors of the user,
In the search session specifying step, the search session is specified so as to include a certain number or less of records indicating that the search session has accessed the search result.
The search result ranking method according to claim 2, wherein: The search log information filtering step includes a search session specifying step for specifying a search session representing a series of search behaviors of the user,
In the search session specifying step, the search session is specified so as to exclude records in which the aggregation key is duplicated.
The search result ranking method according to claim 2, wherein: The search log information filtering step includes a search session specifying step for specifying a search session representing a series of search behaviors of the user,
In the search session specifying step, the search session is specified so as to include a certain number or less of records indicating that the search session has accessed the search result, and to exclude records having the duplicate aggregation key.
The search result ranking method according to claim 2, wherein: The search log information filtering step includes:
A search session filtering step of filtering the search session such that a predetermined time interval exists between the search sessions after the search session specifying process by the search session specifying step;
The search result ranking method according to any one of claims 2 to 5, characterized in that: The search session is
A series of the records in the search action by the same terminal,
Starting from a search request record that is a record indicating a search request in the search action,
The last record in the search log information or the last record of the search request record that appears next is the end point,
The search result ranking method according to any one of claims 3 to 6, characterized in that: The search log information filtering step includes:
A first record filtering step of extracting the records belonging to a specific genre;
The search result ranking method according to any one of claims 2 to 7, characterized in that: The search log information filtering step includes:
A second record filtering step of filtering the records so that a time interval between the records having the same user identifier and the aggregation key is equal to or greater than a predetermined value;
The search result ranking method according to any one of claims 2 to 8, characterized in that: The aggregation key is one or both of a search word and attribute information of the search result.
The search result ranking method according to claim 1, wherein: In the evaluation value tabulation step, the evaluation values are tabulated using the search word and the attribute information of the search result as a key for the number of visits by user tabulated in the user visit count tabulation step.
The search result ranking method according to claim 10. In the evaluation value totaling step, the evaluation values are totaled using the attribute information of the search result as a key for the number of visits by user totaled in the number of revisiting by user.
The search result ranking method according to claim 10. In the evaluation value aggregation step, the average value of the number of visits by user for the key is aggregated as the evaluation value,
The search result ranking method according to claim 11 or 12. In the evaluation value aggregation step, the number of users revisited to the search result is aggregated as the evaluation value.
The search result ranking method according to any one of claims 1 to 13,   The search result ranking method according to any one of claims 1 to 14, wherein, in the evaluation value aggregation step, the aggregation is performed when the number of visits by user is equal to or greater than a certain value.   16. In the evaluation value totaling step, a value obtained by dividing the total evaluation value by the number of unique users for the key is totaled as a final evaluation value. The search result ranking method described. Search log information storage means for storing user search log information;
An evaluation value totaling module that totals evaluation values from the search log information stored in the search log information storage unit;
Score calculation means for calculating scores used for ranking the display order of search results displayed on the search result display screen using the evaluation values tabulated by the evaluation value tabulation module;
Search result display means for performing the ranking using the score and displaying the search results in the order of ranking on the search result display screen;
With
The evaluation value aggregation module
Based on the search log information, based on the number of times the same user identifier and count key have appeared, the number of visits by user totaling means for counting the number of visits by user that is the number of times the user has revisited the search results;
Based on the number of revisits for each user counted by the number of revisits for each user, evaluation value totaling means for totaling the evaluation values;
A search result ranking system comprising: Search log information storage means for storing user search log information;
An evaluation value totaling module that totals evaluation values from the search log information stored in the search log information storage unit;
Score calculation means for calculating scores used for ranking the display order of search results displayed on the search result display screen using the evaluation values tabulated by the evaluation value tabulation module;
Search result display means for performing the ranking using the score and displaying the search results in the order of ranking on the search result display screen;
With
The evaluation value aggregation module
Search log information filtering means for filtering the search log information;
Based on the filtered search log information, based on the number of appearances of the same user identifier and total key, the total number of revisits by user that totals the number of revisits by user, which is the number of times the user has revisited the search results. Means,
Based on the number of revisits for each user counted by the number of revisits for each user, evaluation value totaling means for totaling the evaluation values;
A search result ranking system comprising:

Images