JP2007183859A - Information retrieval system, information retrieval method, and information management apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an information retrieval system, an information retrieval method, and an information management apparatus that can refer to the using history of other user having a similar interest and concern without previously preparing data when a retrieval user retrieves desired information from information stored in a database, and can easily, efficiently, and accurately retrieve information of a further extended content base. <P>SOLUTION: The using history of the retrieval user is compared with that of other user and the similarity is calculated based on the stored using history of each user, the relevance ratio between each information and retrieval condition presented by the retrieval user is calculated based on the using information of the similar users having high similarity, and information of high relevance ratio is retrieved. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an information search system, an information search method, and an information management device for searching information stored in accordance with each user's intention based on search conditions that are accessed and presented by a plurality of users.

  The use of the Internet has become common, and the opportunity for general users to directly search for desired information from a large-scale database such as the Web is increasing. However, searching for a product that suits your preference from a large amount of product data sold on the Web, or searching for a target document or related documents from a large amount of documents such as a Web or in-house document database is an example. Even using a search program takes a lot of time and effort.

  Usually, a keyword search method is often used for such a search program from a large amount of information. However, such low search efficiency is often caused by the fact that the user cannot set an appropriate keyword when using a search program, or the set keyword is ambiguous. The keyword search is a character string search, and even if the range is expanded to synonyms, much information is missing. If the range is widened to prevent this, information different from the desired information will be excessive, and it will be necessary to narrow it down.

  For these keyword-based search challenges, the search that the searcher wants is more about whether the content of the information matches or is similar to the genre, rather than whether the keywords match. Search methods that emphasize categories have been developed (see Patent Document 1).

  In Patent Document 1, the contents are statistically analyzed and classified in advance for each stored information, and the relationship with the corresponding category factor is indexed and stored. At the time of retrieval, the index is referred to. A search technique is proposed in which information strongly related to a category factor corresponding to a search request is referred to. By searching through categories, the accuracy of searching for information that is similar in content to the desired information is improved.

  Also, as a search method that emphasizes the content of information rather than conventional keyword matching, a search that uses the search user's history to find similar information in terms of what information the user has referred to in the past A method has been developed (see Patent Document 2).

  In Patent Document 2, the address of a browsing user's past browsing site is stored, and when searching an Internet site, the information content of the site browsed in the past is decomposed into word levels, and the appearance frequency is statistically determined. Get the analyzed data. On the other hand, a technique has been proposed in which linked sites and related sites are investigated from stored addresses, data analysis is performed, and sites similar to the previously acquired analysis data are found. By searching for similar information based on past reference information, the accuracy of searching for desired information is improved.

  Attempts have also been made to make up for the shortcomings of conventional keyword searches using reference results of other users with similar user attributes, such as the range of interest, gender, and age. For example, in the case of purchasing books through the Internet, the search efficiency is improved by presenting what other users who have purchased the book purchase other books that the user is currently purchasing. At the same time, attempts have been made to realize potential needs.

  Also, not only specific articles and information but also user evaluations, that is, preferences are accumulated as data, and based on comparative evaluation with other users of the data, search for information that is of common interest A technique called “Patent Document 3” has also been developed.

In Patent Document 3, evaluations for each piece of information by each user are accumulated and recorded in advance, and at the time of search, the search user evaluations are compared with those user evaluations, and the similarity of interest is discriminated. A technique has been proposed in which information that is highly evaluated by an evaluator having the same interest and interest is searched. Based on the user evaluation, the accuracy of searching for desired information is improved using the reference results of other users with similar preferences.
JP 2001-184358 A JP 2001-160056 A JP-A-9-190443

  As for the search method from a large amount of information, the conventional keyword search method has various problems as described above. In the technique described in Patent Document 1, although not a character string but a categorical factor is interposed to approach a content-based search, it is only a classification and does not refer to the information content itself. In the technique described in Patent Document 2, the content of the information is not directly grasped, but the content of interest can be interposed indirectly by referring to the past use history. It is based on the data of the reference, and has no expansibility. In the book purchase example, by presenting a purchase book of another user who purchased the same book, it refers to another user who may be interested in the same, but this also intervenes with the book that is about to be purchased It is not possible to say that there is sufficient preference similarity.

  Further, in the technique described in Patent Document 3, similarity of interest is obtained through user evaluation on the contents of information, and similar user evaluation, that is, evaluation by other users having similar interest is used for filtering. However, it is possible to search for desired information in a more expansive manner, but this also requires the user's evaluation for each information to be input in advance.

  The object of the present invention is to solve the above-mentioned problems, and when a search user searches for desired information from information stored in a database, there is no need to prepare data in advance, and other users who have similar interests and interests. It is possible to provide an information search system, an information search method, and an information management device that can refer to the usage history of the user and can easily and more broadly perform content-based information search efficiently and accurately. .

  In order to solve the above problems, the present invention has the following features.

  1. Information storage means for storing information, search condition setting means for setting search conditions for a search user to search for desired information, and the information based on the search conditions set by the search condition setting means An information search unit that searches for information from a storage unit, an information providing unit that selectively uses information from the information searched by the information search unit, and information based on the information providing unit For usage, usage history storage means for storing a usage history for each user including the search user, usage history management means for managing a usage history for each user including the search user stored in the usage history storage means, The usage history management means includes the usage history of the search user stored in the usage history storage means, and the search. Similarity calculation means for calculating the similarity with the use history of other users other than the user, and other than the search user whose use history is similar to the search user based on the similarity calculated by the similarity calculation means Similar user search means for searching for other users, wherein the information search means is information use history of other users other than the search user similar to the use history searched by the similar user search means. Based on the relevance calculation means for calculating the relevance between the information stored in the information storage means and the search condition, and based on the relevance calculated by the relevance calculation means, the information storage means And an associated information retrieval means for retrieving associated information from the information retrieval system.

  2. The information providing means may copy information to a predetermined medium designated by the search user, or copy information to another information storage means different from the information storage means, as a usage form of the selected information. 2. The information search system according to 1, wherein the search user is allowed.

  3. 3. The information search system according to claim 1 or 2, further comprising charging means for charging the search user for use of information by the search user based on the information providing means.

  4). When the usage history of the search user stored in the usage history storage means is less than or equal to a predetermined number of times, the usage history management means calculates the similarity by the similarity calculation means and the similarity user search means 4. The information search system according to any one of claims 1 to 3, wherein a search for similar users based on similarity is not performed and a predetermined other user is set as a similar user with high similarity.

  5. When the usage history of the search user stored in the usage history storage means is a predetermined number or less, the usage history management means sets a virtual usage history for similarity calculation for the search user, The information search system according to any one of claims 1 to 3, wherein the similarity calculation unit calculates the similarity based on the set virtual usage history.

  6). 6. The information retrieval system according to any one of 1 to 5, wherein the usage history storage unit stores a virtual usage history of a virtual user.

  7). 7. The information search according to any one of claims 1 to 6, wherein the similarity calculation means calculates a usage history similarity by excluding other users who have no usage history for a predetermined period of time. system.

  8). The information search system according to any one of claims 1 to 7, wherein the relevance calculation means calculates relevance based on a date and time when information is stored in the information storage means.

  9. 9. The information search system according to any one of 1 to 8, wherein the similarity calculation unit calculates a similarity based on a use date and time in a use history.

  10. A search condition setting step for setting a search condition for a search user to search for desired information, and information search from information storage means in which information is stored based on the search condition set in the search condition setting step An information search step for performing search, an information providing step for selectively using information from the information searched in the information search step, and use of information based on the information providing step. In an information search method comprising: a usage history recording step for storing a usage history for each user including a usage history management step for managing a usage history for each user including the search user stored in the usage history recording step The use history management step includes the search user's use history accumulated in the use history recording step, and the search user. Based on the similarity calculated in the similarity calculation step for calculating the similarity with the usage history of other users outside, and other than the search user whose usage history is similar to the search user based on the similarity calculated in the similarity calculation step A similar user search step of searching for another user, wherein the information search step is based on information use history of other users other than the search user similar to the use history searched in the similar user search step. A relevance calculation step for calculating a relevance between each piece of information stored in the information storage means and the search condition, and a relevance from the information storage means based on the relevance calculated in the relevance calculation step. And a related information search step for searching for information to be performed.

  11. In the information providing step, as a use form of the selected information, copying of information to a predetermined medium designated by the search user or copying of information to another information storage means different from the information storage means is performed. 11. The information search method according to 10, wherein the search user is enabled.

  12 The information search method according to claim 10 or 11, further comprising a charging step of charging the search user for use of information by the search user based on the information providing step.

  13. When the usage history of the search user recorded in the usage history recording step is equal to or less than a predetermined number of times, the usage history management step calculates the similarity in the similarity calculation step and the similarity user search step. 13. The information search method according to any one of 10 to 12, wherein a search for similar users based on similarity is not performed, and a predetermined other user is set as a similar user with high similarity.

  14 When the usage history of the search user recorded in the usage history recording step is less than or equal to a predetermined number of times, the usage history management step sets a virtual usage history for calculating the similarity for the search user, The information search method according to any one of claims 10 to 12, wherein, in the similarity calculation step, the similarity is calculated based on the set virtual use history.

  15. 15. The information search method according to any one of 10 to 14, wherein a virtual usage history of a virtual user is recorded in the usage history recording step.

  16. The information according to any one of claims 10 to 15, wherein, in the similarity calculation step, the similarity of the usage history is calculated by excluding other users who have no usage history for a predetermined period or more. retrieval method.

  17. 17. The information search method according to any one of 10 to 16, wherein in the relevance level calculating step, the relevance level is calculated based on a date and time when information is stored in the information storage unit.

  18. 18. The information search method according to any one of 10 to 17, wherein in the similarity calculation step, the similarity is calculated based on a use date and time in the use history.

  19. Search condition setting means for acquiring and setting search conditions for searching for desired information from a search user, information providing means for selectively using the searched information to the search user, and the information provision Use history storage means for accumulating use history for each user including the search user for use of information based on the means, in the information management apparatus, the use history of the search user stored in the use history storage means A similarity calculation unit that calculates a similarity between the usage history of other users other than the search user, and a similarity between the search user and the usage history based on the similarity calculated by the similarity calculation unit Similar user search means for searching for other users other than the search user, and for the information storage means for storing information, the search condition setting means Based on the set search condition and the information use history of other users other than the search user similar to the use history searched by the similar user search means, the information search is instructed and the search result is acquired. An information management apparatus characterized by that.

  According to the present invention, when a search user searches for desired information from information stored in a database, the use history of each stored user is compared with the use history of the search user and other users. And a similar user having a high similarity can be obtained. Further, based on the usage information, the degree of association between each piece of information and the retrieval condition presented by the retrieval user can be calculated, and information with a high degree of association can be retrieved.

  This makes it possible to refer to the usage history of other users with similar interests without having to prepare data in advance, making it easier and more efficient to search information based on content more easily. In addition, it is possible to provide an information search system, an information search method, and an information management device that can be performed with high accuracy.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

(Configuration of information retrieval system 1)
First, the configuration of an information search system according to an embodiment of the present invention will be described with reference to FIG. FIG. 1 shows a configuration example of the information search system 1.

  In this system configuration example, information to be searched for and used is referred to as content, and many users access the information search system 1 having a large amount of content. Each user presents a search condition, requests use of the searched content, and uses the content by responding to a charge from the system administrator.

  The content can be in any form other than documents, such as music data, images, and moving images. Depending on the form of content, the form of use is also different. When dealing with merchandise, the content is information about the merchandise, but the user obtains the merchandise by responding to the charge.

  In FIG. 1, 11 is a content server, which stores content as information, performs search processing upon request, and provides content through a network. A user information management server 12 stores and manages a user who uses the content and information regarding the usage history of the content. A billing server 13 performs billing and settlement processing for the use of content by the user. A user terminal 15 is used by the user to perform input operations for content use, display for content use, printing, and the like. Reference numerals 16 and 17 denote terminals of other users.

  The content server 11, user information management server, and billing server installed by the content provider as the information search system 1 are connected to the user terminal 15, 16 or 17 via the network 14, and each user Can be searched and used by accessing each user terminal 15 via the network.

  The content server 11 stores content as information, a content database (content DB) 111 that functions as information storage means, a search processing unit 112 that functions as information search means, searches information from the content DB, and a network And a network connection unit 113 for transmitting and receiving requests and instructions related to search and use.

  In addition, the search processing unit 112 receives a search condition such as a search key, and searches for and extracts a relevance level calculation unit 112a that calculates a relevance level with each information content, and a related information content based on the relevance level. And a related information search unit 112b. Accordingly, the relevance calculation unit 112a functions as a relevance calculation unit, and the related information search unit 112b functions as a related information search unit.

  The user information management server 12 is an information management apparatus that stores information about users and manages a user management database (user management DB) 121 that functions as a usage history storage unit for information contents and a user information content usage history. , A usage history management unit 122 that functions as a usage history management unit, a user management unit 123 that authenticates a user and controls and manages content usage for the user, and transmits and receives requests and instructions related to search and usage via a network And a network connection unit 124.

  The usage history management unit 122 refers to the usage history of the search user, calculates the similarity with other users' usage history, and another user whose usage history is similar based on the similarity And a similar user search unit 122b for searching for and extracting. Therefore, the similarity calculation unit 122a functions as a similarity calculation unit, and the similar user search unit 122b functions as a similar user search unit.

  The user management unit 123, in cooperation with the user terminal 15 described later, accepts access by the user and acquires conditions related to the search such as information about the user and a search key. Therefore, it functions as search condition setting means. In addition, the user management unit 123 performs processing in response to the user's content use request and outputs a content use permission. Therefore, it also functions as information providing means.

  The billing server 13 functions as a billing unit, and billing processing unit 132 that performs billing processing in response to a user's content usage request, settlement processing unit 131 that performs settlement processing in response to a user response to billing processing, and billing And a network connection unit 133 for transmitting and receiving requests and instructions related to.

  The user terminal 15 is a terminal connected to a network, such as a personal computer, for example, a network connection unit 155 for accessing the information search system 1, an input unit 151 for performing an input operation, and a display unit for displaying information. 152, a printing unit 153 for printing information on a medium, a storage unit 154 for storing information, and a control unit 156 for controlling operations thereof.

(Operation of information retrieval system 1)
FIG. 2 is a sequence diagram showing the operation of the information search system 1 according to this embodiment. The operation of the information search system 1 will be described with reference to FIG. Reference is also made to FIGS. 1 and 3 as appropriate.

  FIG. 2 shows the order and flow of the operations of the user terminal 15, the content server 11, the billing server 13, and the user information management server 12 in time series from the top to the bottom of the drawing. It demonstrates along the sequence.

  When the user accesses the information search system 1 from the user terminal 15, for example, the user information, information for specifying the user such as the user ID number as the search information 211, and conditions for searching for the desired information content such as the search key are stored It transmits to the management server 12.

  In the user information management server 12, user authentication 221 is performed by the user management unit 123, and the usage history reference 222 of the corresponding user is performed by the usage history management unit 122. In either case, the user management DB 121 is referred to.

  FIG. 3A shows the internal data structure of the user management DB 121. The user authentication 221 refers to user data 41 (including an ID number, a password, and other personal information). The user data 41 is also referred to for billing processing described later. In the usage history reference 222, the usage history 42 is referred to. The usage history 42 includes the ID number of each content used in addition to the user ID, and information such as the usage date, billing information, and usage mode for each content. The following similar user search process is performed using these usage history data.

  Subsequently, a similar user search 223 is performed in the usage history management unit 122, and by comparing with the usage history of other users, extraction of N users (hereinafter also referred to as N users) with high similarity, that is, N users Extraction 224 is performed, and N user use content information 225 is transmitted to the content server 11. The similar user search process will be described later in detail.

  In the content server 11, after the search processing unit 112 performs the weight setting 231 based on the N user use content information 225, the related content search 232 using the search condition presented by the user has a high degree of relevance. M content extraction 233 is performed and transmitted to the user terminal 15 for presentation to the user. For related content search processing, the content DB 111 is referred to.

  FIG. 3B shows the internal data structure of the content DB 111. Here, a book is shown as an example of content. The book content 31 is obtained by converting the content of the book itself as information into data, and is a content body provided to the user. The search metadata 32 is information relating to content and is data used for search. Specifically, the ID number, title, author name, publisher issue date, classification, keyword, and the like for each book are searched according to search conditions designated by the user. For example, if a search key (consisting of a plurality of words) is designated by the user, the search process is performed by matching with a keyword added according to the contents of the book.

  The related content search process will be described later in detail.

  In the user terminal 15, M content having a high degree of relevance is displayed on the display unit 152, and the user selects a desired content using the input unit 151, and transmits a content designation / use request 212 to the user information management server 12.

  In the user information management server 12, the user management unit 123 receives a usage request from the user and transmits a charging processing request 226 to the charging server 13.

  In the billing server 13, billing processing 241 is performed in the billing processing unit 132, and billing details are transmitted to the user terminal 15.

  In the user terminal 15, the charging contents are displayed on the display unit 152, and the user transmits the payment permission and information for payment to the charging server 13 as the payment process 213 using the input unit 151.

  The billing server 13 receives the payment process 213, performs a settlement process 242 for the billing, and transmits the settlement process information to the user information management server 12.

  In the user information management server 12, the user management unit 123 receives the billing settlement 242 and transmits a content provision permission 227 to the content server 11.

  In response to the content provision permission 227, the content server 11 enables the content to be provided to the user, and transmits the content provision 234 to the user terminal 15.

  In the user terminal 15, the content provision 234 is displayed on the display unit 152, and the user uses the provided content using the input unit 151. Specifically, there are usage forms such as downloading contents and storing them in the storage unit 154, printing them with the printing unit 153, or displaying them on the display unit 152.

  On the other hand, in the user information management server 12, the usage history management unit 122 accesses the user management DB 121 and updates the usage history 42 of the corresponding user. That is, information such as the ID number and date of newly used content, billing information, and usage form is added.

(Overall processing flow)
The information search method in the information search system 1, that is, the processing flow of the entire content search and use flow will be described with reference to FIG. FIG. 4 is a flowchart showing the flow of content search and use.

  First, in step S1, when a user accesses the information search system 1 from the user terminal 15 and a condition for searching for desired information content such as a search key is input by the user, user management of the user information management server 12 is performed. The unit 123 acquires and sets a condition for searching for desired information content such as a search key together with information specifying the user such as a user ID number. That is, step S1 functions as a search condition setting step.

  Next, in step S2, the user management unit 123 of the user information management server 12 refers to the user data 41 of the user management DB 121 based on the user ID number and password, and performs user authentication. In addition, the usage history management unit 122 refers to the user usage history 42 and uses it as a reference for the similar user search process to be performed next.

  In step S <b> 3, the usage history management unit 122 refers to the user management DB 121, calculates the similarity of the usage history of other users with respect to the usage history of the search requesting user, and performs a similar user detection process. That is, step S3 functions as a usage history management process. Details of the step S3 will be described later with reference to the flowchart of the similar user search process of FIG.

  In this embodiment, the process of step S3 is executed every time a search is performed. However, the process is periodically executed every predetermined time, and another user having a high usage history similarity is obtained for each user. You may keep it.

  In step S4, the content server 11 receives used content information of N users with high similarity, the search processing unit 112 sets weights based on the information, refers to the content DB 111, and calculates the degree of association between the search key and each content. Perform related content search processing. That is, step S4 functions as an information search process. Details of the step S4 will be described later with reference to the flowchart of related content search processing of FIG.

  In step S5, M content having a high degree of relevance is displayed on the display unit 152 of the user terminal 15, and the user selects and designates desired content using the input unit 151, and makes a use request to the user management server. The user management unit 123 receives a usage request from the user and transmits a charging process request to the charging server 13.

  In step S 6, billing processing is performed in the billing processing unit 132 of the billing server 13, and billing information is transmitted to the user terminal 15. In step S7, the contents of the charge are displayed on the display unit 152 of the user terminal 15, and when the user uses the input unit 151 to send the payment permission and information for payment to the charge server 13 as a payment process, The billing server 13 performs billing settlement processing in response to the payment processing, and transmits settlement processing information to the user information management server 12. That is, step S6 and step S7 function as a charging process.

  In step S8, the user management unit 123 of the user information management server 12 receives the billing settlement information and transmits a content provision permission to the content server 11, and the content server 11 can receive the content provision permission and provide the content to the user. The provision of the corresponding content is transmitted to the user terminal 15. That is, step S8 functions as an information providing process. In the user terminal 15, content provision is displayed on the display unit 152, and the user displays the provided content on the display unit 152 using the input unit 151, prints it with the printing unit 153, downloads it, and stores it in the storage unit 154. It is used in the usage form such as.

  In step S9, the usage history management unit 122 of the user information management server 12 accesses the user management DB 121 and updates the usage history of the user who used the content. That is, information such as the ID number and date of the newly used content, billing information, and usage mode is added to the usage history of the user management DB 121. That is, step S9 functions as a usage history recording process.

(Similar user search processing example 1)
FIG. 5 is a detailed flowchart of the step S3 of FIG. 4, that is, the similar user search process. A flow of the similar user search process will be described with reference to FIG.

  In step S31, similarity calculation processing is performed by the similarity calculation unit 122a. That is, the similarity of the usage history is calculated for each user, and the usage history is ordered. Step S31 functions as a similarity calculation step.

  In step S32, the similar user search unit 122b searches for and extracts N users with high similarity based on the similarity of each user to the usage history of the search requesting user. Step S32 functions as a similar user search step.

  Before describing the similarity calculation step, the similarity will be described first.

  Let p be the search requesting user and q be any other user. Any user can refer to the use history by the user management DB 121.

  In order to define the similarity of the content usage history of any other user q with respect to the content usage history of the user p, first, a similar condition for an arbitrary content i is defined as follows.

  Similar conditions: Arbitrary content i is used by user p and also used by other users q.

For content that satisfies this condition, the value is increased for each content, and s is determined for each content.
When content i satisfies the above similar condition for user q, si (p, q) = 1,
Si (p, q) = 0 when the content i does not satisfy the above similar condition for the user q
Si (p, q) is defined so that (see Table 1).

  In Table 1, for any content i for any other user q, each si (p, q) has a value of 1 or 0 according to the above definition.

  By taking the sum of si (p, q) for all the contents, the similarity S (p, q, so that the numerical value increases as the number of contents satisfying the similar condition increases, that is, the similarity increases. q) is defined. That is,

  The similarity S (p, q) is calculated for all other users q according to the above equation (1). Thereby, the user can be ordered according to the similarity of the usage history.

  In this embodiment, in order to obtain N users with high similarity, similarity is calculated for all other users. However, by building a search index structure in advance, It is also possible to obtain efficiently.

(Similarity calculation processing flow)
Next, an example of a specific similarity calculation procedure will be described.

  FIG. 6 is a detailed flowchart of step S31 of FIG. 5, that is, a step of similarity calculation processing. A flow of similarity calculation processing will be described with reference to FIG.

  In step S310, the use history of the search requesting user p is referred to as the similarity criterion. This is a reference for calculating the similarity of the usage history of other users.

  In step S311, any other unset user q is set as a user who compares to obtain the similarity. The processes from here to step S317 are repeated for all other users while changing the setting of the user q. That is, the similarity S (p, q) is calculated for all other users.

  In step S312, the similarity S (p, q) is set to 0 as an initial setting for the set user q. By repeating the addition of si for all contents with respect to this S (p, q), the similarity S (p, q) represented by the equation (1) is calculated.

  In step S313, an arbitrary content i that has not been set is set as the reference content for similarity calculation. The processes from here to step S315 are repeated for all contents by changing the setting of the contents i. That is, si (p, q) is calculated depending on whether similar conditions are satisfied for all contents (step S314). Further, for all contents, si (p, q) of the referenced content i is added to the similarity S (p, q) (step S315), and as a result, the repetition for all the contents is completed. The similarity S (p, q) represented by (1) as the sum of si (p, q) is calculated (step S317).

  In step S316, it is determined whether there is no unset content, that is, whether the processing has been completed for all content. When the process is completed (step S316: YES), step S317 is executed. If not completed (step S316: NO), the process returns to step S313, sets unset content, and repeats the processing up to step S315.

  In step S317, the similarity S (p, q) calculated for the user q is temporarily stored for later comparison and ranking.

  In step S318, it is determined whether there are no other users who have not been set, that is, whether the processing has been completed for all other users. When the process is completed (step S318: YES), step S319 is executed. If not completed (step S318: NO), the process returns to step S311 to set another user who has not been set and repeat the processing up to step S317.

  In step S319, similarities S (p, q) for all other users are prepared, and the temporarily stored similarities are compared, and the other users are ordered according to the level of similarity.

  Thereby, N users with high similarity in step S32 in FIG. 5 can be extracted, and the similar user search process can be completed.

(Related Content Search Processing Example 1)
FIG. 7 is a detailed flowchart of step S4 of FIG. 4, that is, the related content search process. The flow of related content search processing will be described with reference to FIG.

  In step S41, the relevance calculation unit 112a performs relevance calculation processing. That is, the degree of association with the search condition is calculated for each content, and the order is based on the degree of association. Step S41 functions as an association degree calculation step.

  In step S42, the related information search unit 112b searches for and extracts M contents having a high degree of association based on the degree of association of each content with respect to the search condition designated by the search requesting user. Step S42 functions as a related information search process.

  Before describing the relevance level calculation step, the relevance level will be described first.

  The search condition specified by the search requesting user is assumed to be a search key composed of a plurality of words. As described above, search metadata is assigned to each content, and the search metadata can be referred to by the content DB 111. In the present embodiment, the degree of association is calculated using a keyword as search metadata.

  However, it can be easily extended to metadata other than keywords by a similar method. The degree of relevance may be the full text text of the document instead of the metadata. If an image is included in the document, the relation with the image may be calculated and added to the relevance calculation.

  In order to define the degree of association between the search key designated by the user p and the arbitrary content i, first, the relationship condition for the arbitrary keyword k is determined as follows.

  Related conditions: An arbitrary keyword k is included in the search key and also included in the metadata of the content i.

  In this embodiment, the keyword is included in the metadata of the content, but a search method in which the content itself is a document and the keyword is included in the content itself may be used. Considering this, hereinafter, the expression that the keyword is included in the content or appears in the content is used, including the case where the keyword is included in the metadata.

  This related condition is divided into the first condition a and the second condition b as follows.

Condition a: Arbitrary keyword k is included in the search key Condition b: Arbitrary keyword k is included in the content i First, the numerical value increases when this condition a is satisfied, and decreases otherwise. X (k) is determined for each keyword, that is,
When the keyword k satisfies the condition a, x (k) = 1,
When the keyword k does not satisfy the condition a, x (k) = 0,
X (k) is defined such that (see Table 2).

  In Table 2, for any keyword k, each x (k) has a value of 1 or 0 according to the above definition.

Further, d (i, k) that increases when the above condition b is satisfied, and decreases otherwise is obtained for a combination of an arbitrary content i and an arbitrary keyword k.
When content i and keyword k satisfy the above condition b, d (i, k) = 1,
When content i and keyword k do not satisfy the above condition b, d (i, k) = 0,
D (i, k) is defined so that (see Table 3).

  In Table 3, for a combination of an arbitrary keyword k and an arbitrary content i, each d (i, k) has a value of 1 or 0 according to the above definition.

  Here, when x (k) × d (i, k) is obtained for the combination of the content i and the keyword k, the condition a and the condition b are the same for the x (k) × d (i, k). If the relation is satisfied, that is, the relation condition is satisfied, the numerical value becomes large (x (k) × d (i, k) = 1), and otherwise, the numerical value becomes small (x (k) × d (i, k) = 0). ).

  By taking the sum of x (k) x d (i, k) for all keywords, the number increases as the number of keywords that satisfy the related condition increases, that is, the degree of relevance increases. A degree R (i) is defined. That is,

In equation (2), w (k) is
w (k) = 1 + α · u (k) / v (k) (3)
It is.

  w (k) is a coefficient that becomes a weight when the above-mentioned x (k) × d (i, k) is summed up for the keyword to obtain the relevance R (i). See Table 4 for u (k) / v (k) in the weight calculation formula (3).

  In Table 4, u (k) is the appearance frequency of the keyword k in all contents used by N users with high similarity, and v (k) is the appearance frequency of the keyword k in all contents. Therefore, u (k) / v (k) in the expression (3) increases the relevance when a keyword that appears frequently in content used by a user with a high degree of similarity appears in the content. It is a serious factor. As a result, it is possible to calculate a more appropriate relevance level of the content using the use history of the user having a high degree of similarity.

  The higher the appearance frequency of the content used by users with a high degree of similarity, the higher the degree of relevance, but the degree of effect is controlled by the weighting factor α for u (k) / v (k). . Further, 1 in the expression (3) is a term for enabling a search even when no keyword appears in the content used by a user having a high degree of similarity.

  Relevance R (i) is calculated for all contents i according to the above equation (2). As a result, the contents can be ordered by the degree of association with the search key.

  In this embodiment, in order to obtain M pieces of content having a high degree of relevance, the degree of relevance is calculated for all contents. However, by building a search index structure in advance, It is also possible to obtain efficiently.

(Relevance calculation process flow)
Next, an example of a specific relevance calculation procedure will be described.

  FIG. 8 is a detailed flowchart of step S41 of FIG. 7, that is, a step of relevance calculation processing. With reference to FIG. 8, the flow of relevance calculation processing will be described.

  In step S410, a search condition designated by the search request user p, here a search key including a plurality of words, is referred to as a criterion for the degree of association. This is a reference for calculating the relevance of any content.

  In step S411, arbitrary unset content i is set as the content for which the relevance is obtained by comparison. The processes from here to step S417 are repeated for all contents by changing the setting of the contents i. That is, the relevance R (i) is calculated for all contents.

  In step S412, the degree of association R (i) is set to 0 as an initial setting for the set content i. For this R (i), by repeating the addition of x (k) × d (i, k) multiplied by the weighting factor w (k) for all keywords, The degree of association R (i) represented is calculated.

  In step S413, an arbitrary keyword k that has not been set is set as a reference keyword for calculating the degree of association. The processes from here to step S415 are repeated for all the keywords while changing the setting of the keyword k. That is, x (k), d (i, k), and weighting factor w (k) are calculated depending on whether the relevant conditions are satisfied for all keywords (step S414). In addition, for all keywords, the relevance R (i) is multiplied by x (k) × d (i, k) for the referenced keyword k multiplied by a weighting factor w (k) (step) S415) As a result, when the repetition for all keywords is completed, the degree of relevance as the sum of x (k) × d (i, k) multiplied by the weighting factor w (k) represented by equation (2) R (i) is calculated (step S417).

  In step S416, it is determined whether there are no unset keywords, that is, whether the processing has been completed for all keywords. When the process is completed (step S416: YES), step S417 is executed. If not completed (step S416: NO), the process returns to step S413, an unset keyword is set, and the process up to step S415 is repeated.

  In step S417, the degree of association R (i) calculated for the content i is temporarily stored for later comparison and ranking.

  In step S418, it is determined whether there is no unset content, that is, whether processing has been completed for all content. If it has been completed (step S418: YES), step S419 is executed. If not completed (step S418: NO), the process returns to step S411, sets unset content, and repeats the processing up to step S417.

  In step S419, the relevance levels R (i) for all the contents are gathered, the temporarily stored relevance levels are compared, and the contents are ordered according to the level of relevance.

  Accordingly, the M content having a high degree of relevance in step S42 in FIG. 7 can be extracted, and the related content search process can be completed.

(Similar user search processing example 2)
FIG. 9 is a flowchart illustrating another example of similar user search processing described with reference to FIG. A similar user search processing example 2 will be described with reference to FIG.

  In step S51, it is determined whether or not the usage history of the referred search user is sufficient to calculate the similarity. For example, the determination is performed based on whether the number of times of use or the number of used contents is equal to or greater than a predetermined threshold. If it is equal to or greater than the predetermined threshold (step S51: YES), steps S31 and S32 are executed. Steps S31 and S32 are the same as those already described with reference to FIG.

  If the number of times of use or the number of used contents is not equal to or greater than a predetermined threshold (step S51: NO), step S52 is executed, and the normal steps S31 and S32 are not executed. That is, in step S52, a predetermined user set in advance is set as a user having a high degree of similarity. Alternatively, a virtual user having a virtual usage history may be set.

  In this way, even when the search user's usage history is not sufficient, a representative user or a virtual user is determined in advance by using the user's profile or assuming a specific tendency, and similar users. By setting as, the related content can be searched more appropriately.

(Similar user search processing example 3)
FIG. 10 is a flowchart showing another example of similar user search processing described with reference to FIG. A similar user search processing example 3 will be described with reference to FIG.

  In step S51, it is determined whether or not the usage history of the referred search user is sufficient to calculate the similarity. For example, the determination is performed based on whether the number of times of use or the number of used contents is equal to or greater than a predetermined threshold. If it is equal to or greater than the predetermined threshold (step S51: YES), steps S31 and S32 are executed. Steps S31 and S32 are the same as those already described with reference to FIG.

  When the number of times of use or the number of contents used is not equal to or greater than a predetermined threshold (step S51: NO), step S53 is executed, and then the process returns to the normal execution of steps S31 and S32. In step S53, a predetermined predetermined usage history is set as a virtual usage history of the search user. Alternatively, the user may be caused to input a virtual usage history by a questionnaire method or the like. Based on this virtual usage history, similarity calculation in the next step S31 is performed.

  In this way, even when the search user's usage history is not sufficient, by searching for a user with similar preferences by setting a virtual usage history or by allowing the user to input, the user can be a similar user. It is possible to search for related contents, and it is possible to search for similar users with higher accuracy and user searches with high relevance.

(Similar user search processing example 4)
FIG. 11 is a flowchart illustrating another example of the similar user search process described with reference to FIG. A similar user search processing example 4 will be described with reference to FIG.

  In step S54, it is determined whether or not the predetermined setting is desired to be searched higher. For example, if the administrator of the information search system 1 desires that a predetermined content is searched for higher, the predetermined content is searched higher by registering a virtual user having a virtual usage history. Can also be guided.

  If it is a setting which desires a high-order search by step S54 (step S54: YES), step S55 will be performed. A step S55 registers a virtual user who has a history that the content for which a higher search is desired is virtually used together with the same type of content. Thereafter, by returning to the execution of the normal steps S31 and S32, the registered virtual user can be easily searched as a similar user, and the possibility that a predetermined content is searched as related content also increases. .

  In step S54, if it is not a setting that desires a higher search (step S54: NO), the normal steps S31 and S32 are executed as they are. Steps S31 and S32 are the same as those already described with reference to FIG.

  In this way, even when it is desired to search for predetermined content that is determined in advance, by registering a virtual user having a virtual usage history, the registered virtual user is searched as a similar user. As a result, the possibility that a predetermined content is searched as related content can be increased.

(Similar user search processing example 5)
FIG. 12 is a flowchart illustrating another example of similar user search processing described with reference to FIG. A similar user search processing example 5 will be described with reference to FIG.

  In step S56, for other users to be searched, it is determined whether or not to search for an old user having a usage history that is too old as data for calculating the similarity. If an old user is excluded (step S56: YES), step S57 is executed. If the old user is not excluded (step S56: NO), steps S31 and S32 are executed as usual. Steps S31 and S32 are the same as those already described with reference to FIG.

  In step S57, a setting is made to exclude users who have not been used for a predetermined period of time from other users to be searched. Alternatively, a predetermined period may be input and set. For example, referring to the latest use time, a user who has not accessed after that exceeds a predetermined period is set as an old user to be excluded from the search. In other words, only users who have access during the set period are set as target users after the next similarity calculation process.

  Next, step S31 and step S32 are executed. Steps S31 and S32 differ from the case of FIG. 5 described above in that the old user and the set user are excluded from the search. The rest is the same, and the description is omitted.

  In this way, even if the usage history of other users to be searched is old and the usage history is not very helpful even if it is a similar user, in the search for related content, it is excluded from the search target. Therefore, it is possible to prevent the possibility that content that is too old will be searched, and to search for the latest content with a high degree of relevance through the search for the latest similar user.

(Related content search processing example 2)
FIG. 13 is a flowchart illustrating another example of the related content search process described with reference to FIG. The related content search processing example 2 will be described with reference to FIG.

  In step S61, it is determined whether or not new content is regarded as more important as content with a high degree of relevance. If new content is to be used heavily (step S61: YES), step S62 is executed. When that is not right (step S61: NO), step S41 and step S42 are performed as usual. Steps S41 and S42 are the same as those already described with reference to FIG.

  In step S62, the calculation formula of the weight coefficient w (k) used for the next relevance calculation process is changed. For example, the weighting coefficient w (k) is a calculation formula that varies depending on the registration time of the content i as in the following formula (4).

That is, w (k) in equation (2) is
w (k, i) = 1 + α · u (k) / v (k) + β · 1 / ΔT (i) (4)
Is replaced by

  Here, ΔT (i) indicates a period from the registration of the content i, and β is a predetermined constant indicating the degree of weighting for the period.

  After changing the weighting coefficient w (k) in this way, the following steps S41 and S42 are executed. Steps S41 and S42 differ from the case of FIG. 7 described above only in the weight coefficient w (k) used for the above-described relevance calculation. The rest is the same, and the description is omitted.

  That is, in the relevance calculation process in step S41, by using the weighting coefficient w (k, i) of Expression (4), the weighting coefficient increases for newly registered new contents, and the relevance degree increases for new contents. Will be.

  In this way, when content to be searched includes content that is too old to be used for reference, content that is too old is searched for by searching for related content by changing the weighting factor according to the newness of the content. The possibility that the latest content is reduced, the latest content can be easily searched, and the latest content search with high relevance can be performed.

(Similar user search processing example 6)
FIG. 14 is a flowchart illustrating another example of the similar user search process described with reference to FIG. A similar user search processing example 6 will be described with reference to FIG.

  In step S58, for searching for similar users, it is determined whether or not the user's recent preference, that is, a new usage history is more important as data for calculating similarity. If the new usage history is to be used heavily (step S58: YES), step S59 is executed. When that is not right (step S58: NO), step S31 and step S32 are performed as usual. Steps S31 and S32 are the same as those already described with reference to FIG.

  In step S59, the definition of si (p, q) used for the next similarity calculation process is changed. For example, si (p, q) is defined to be different depending on the use time of the content i, not necessarily 1, even when similar conditions are satisfied as follows.

When content i satisfies the similar condition described above for user q
si (p, q) = 1 / ΔT (i),
When content i does not satisfy the similar condition described above for user q
si (p, q) = 0
It is defined as

  Here, ΔT (i) indicates a period from the latest use date and time of the content i. The latest usage date and time is calculated from the usage history of the user p or the user q.

  After changing si (p, q) in this way, the following steps S31 and S32 are executed. Steps S31 and S32 differ from the case of FIG. 5 described above only in si (p, q) used for the above-described similarity calculation. The rest is the same, and the description is omitted.

  That is, in the similarity calculation process of step S31, the recent user's preference is calculated by using the changed si (p, q), with emphasis on the recent usage of the content in the similarity calculation. Will be reflected.

  In this way, when the usage history used for the similar user search includes an old usage history, that is, content usage reflecting the user's old preference, the similarity is calculated according to the new usage timing of each content. By changing the definition of, the possibility of searching for similar users based on the user's past usage history is reduced in the search for similar users, making it easier to perform searches based on the latest usage history and the latest usage. A similar user search having a high similarity in the history can be performed.

  As described above, according to the present embodiment, when the search user searches for desired information from the information stored in the database, the use history of the search user and other users is determined from the stored use history for each user. Can be calculated by calculating the similarity, and a similar user having a high similarity can be obtained. Further, based on the usage information, the degree of association between each piece of information and the retrieval condition presented by the retrieval user can be calculated, and information with a high degree of association can be retrieved. This makes it possible to refer to the usage history of other users with similar interests without having to prepare data in advance, making it easier and more efficient to search information based on content more easily. In addition, it is possible to provide an information search system, an information search method, and an information management device that can be performed with high accuracy.

  The embodiment of the present invention is not limited to the above-described embodiment, and various modified forms are also included in the scope thereof as long as the gist of the present invention is met.

It is a block diagram which shows the structural example of the information search system 1 which concerns on this embodiment. It is a sequence diagram which shows the operation example of the information search system 1 which concerns on this embodiment. It is a figure which shows the example of the internal data structure of the user management DB121 in FIG. 1, and content DB111. 3 is a flowchart showing an overall flow regarding content search and use in the information search system 1. FIG. 5 is a flowchart for explaining step S <b> 3 in FIG. 4, that is, a similar user search process. 6 is a detailed flowchart of step S31 in FIG. 5, that is, a step of similarity calculation processing. FIG. 5 is a flowchart illustrating step S4 in FIG. 4, that is, a process of related content search processing. It is a detailed flowchart of the process of step S41 in FIG. It is a flowchart which shows another process example 2 of the similar user search process in FIG. It is a flowchart which shows another process example 3 of the similar user search process in FIG. It is a flowchart which shows another process example 4 of the similar user search process in FIG. It is a flowchart which shows another process example 5 of the similar user search process in FIG. It is a flowchart which shows another process example 2 of the related content search process in FIG. It is a flowchart which shows another process example 6 of the similar user search process in FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Information search system 11 Content server 12 User information management server 13 Billing server 14 Computer network 15, 16, 17 User terminal 31 Content 32 Search metadata 41 User data 42 Usage history 111 Content DB
112 Search processing unit 112a Relevance calculation unit 112b Related information search unit 113, 124, 133, 155 Network connection unit 121 User management DB
122 Usage history management unit 122a Similarity calculation unit 122b Similar user search unit 123 User management unit 131 Settlement processing unit 132 Charge processing unit 151 Input unit 152 Display unit 153 Printing unit 154 Storage unit 156 Control unit

Claims (19)

Information storage means for storing information; and
A search condition setting means for setting a search condition for a search user to search for desired information;
Information search means for searching for information from the information storage means based on the search conditions set by the search condition setting means;
Information providing means for selectively using information from the information searched by the information searching means to the search user;
Regarding the use of information based on the information providing means, usage history storage means for storing a usage history for each user including the search user;
In an information search system comprising: a usage history management unit that manages a usage history for each user including the search user stored in the usage history storage unit;
The usage history management means includes:
Similarity calculation means for calculating the similarity between the usage history of the search user stored in the usage history storage means and the usage history of other users other than the search user;
Similar user search means for searching for other users other than the search user whose use history is similar to the search user based on the similarity calculated by the similarity calculation means;
The information search means includes
Based on the information use history of other users other than the search user similar in the use history searched by the similar user search means, the degree of association between each information stored in the information storage means and the search condition is determined. Relevance calculating means for calculating;
Related information search means for searching related information from the information storage means based on the relevance calculated by the relevance calculation means;
An information retrieval system characterized by that. The information providing means includes
The search user can copy the information to a predetermined medium designated by the search user or copy the information to another information storage means different from the information storage means as the use form of the selected information. Let
The information search system according to claim 1. Charging means for charging the search user for use of information by the search user based on the information providing means;
The information search system according to claim 1 or 2, characterized in that When the usage history of the search user stored in the usage history storage means is a predetermined number of times or less,
The usage history management means includes:
Without calculating the similarity by the similarity calculation means and searching for similar users based on the similarity by the similar user search means,
Set a predetermined other user as a similar user with a high similarity,
The information search system according to any one of claims 1 to 3, wherein When the usage history of the search user stored in the usage history storage means is a predetermined number of times or less,
The usage history management means sets a virtual usage history for similarity calculation for the search user,
The similarity calculation means calculates the similarity based on the set virtual usage history;
The information search system according to any one of claims 1 to 3, wherein The usage history storage means includes
The virtual user's virtual usage history is stored,
The information search system according to claim 1, wherein the information search system is an information search system. The similarity calculation means includes:
Exclude other users who have no usage history for a certain period of time and calculate the similarity of usage history.
The information search system according to claim 1, wherein the information search system is an information search system. The relevance calculation means includes:
Calculating the degree of relevance based on the date and time when the information was stored in the information storage means;
The information search system according to claim 1, wherein the information search system is an information search system. The similarity calculation means includes:
Calculate similarity based on usage date and time in usage history,
The information search system according to claim 1, wherein the information search system is an information search system. A search condition setting step for setting a search condition for the search user to search for desired information;
Based on the search condition set in the search condition setting step, an information search step of searching for information from the information storage means in which information is stored;
An information providing step for selectively using information to the search user from the information searched in the information search step;
Regarding the use of information based on the information providing step, a usage history recording step for accumulating a usage history for each user including the search user;
A usage history management step of managing a usage history for each user including the search user accumulated in the usage history recording step;
The usage history management step includes:
A similarity calculation step of calculating a similarity between the usage history of the search user and the usage history of other users other than the search user, accumulated in the usage history recording step;
A similar user search step of searching for other users other than the search user similar to the search user based on the similarity calculated in the similarity calculation step,
The information search step includes
Based on the information use history of other users other than the search user similar to the use history searched in the similar user search step, the degree of association between each information stored in the information storage means and the search condition is determined. Relevance calculation step to calculate,
A related information search step of searching for related information from the information storage means based on the relevance calculated in the relevance calculation step.
An information search method characterized by that. In the information providing step,
As the usage form of the selected information, the search user can copy information to a predetermined medium designated by the search user or copy information to another information storage means different from the information storage means. Become,
The information search method according to claim 10. A charging step of charging the search user for use of information by the search user based on the information providing step;
The information search method according to claim 10 or 11, wherein When the usage history of the search user recorded in the usage history recording step is a predetermined number of times or less,
In the usage history management step,
Calculation of similarity in the similarity calculation step and search for similar users based on the similarity in the similar user search step are not performed,
A predetermined other user is set as a similar user having a high degree of similarity.
The information search method according to claim 10, wherein the information search method is a data search method. When the usage history of the search user recorded in the usage history recording step is a predetermined number of times or less,
In the usage history management step, a virtual usage history for similarity calculation is set for the search user,
In the similarity calculation step, the similarity is calculated based on the set virtual usage history.
The information search method according to claim 10, wherein the information search method is a data search method. In the usage history recording step,
The virtual usage history of the virtual user is recorded,
The information search method according to claim 10, wherein the information search method is a data search method. In the similarity calculation step,
The usage history similarity is calculated by excluding other users who have no usage history for a predetermined period of time.
The information search method according to any one of claims 10 to 15, wherein In the relevance calculation step,
The relevance is calculated based on the date and time when the information is stored in the information storage unit.
The information search method according to any one of claims 10 to 16, wherein In the similarity calculation step,
The similarity is calculated based on the usage date and time in the usage history.
The information search method according to any one of claims 10 to 17, wherein: Search condition setting means for acquiring and setting search conditions for searching for desired information from a search user;
Information providing means for selectively using the searched information to the search user;
Regarding the use of information based on the information providing means, usage history storage means for storing a usage history for each user including the search user;
In an information management device having
Similarity calculation means for calculating the similarity between the usage history of the search user stored in the usage history storage means and the usage history of other users other than the search user;
Similar user search means for searching for other users other than the search user whose use history is similar to the search user based on the similarity calculated by the similarity calculation means;
For the information storage means in which the information is stored, the search conditions set by the search condition setting means and the other users other than the search user similar in the usage history searched by the similar user search means Based on the information usage history, instruct information search and obtain the search results,
An information management apparatus characterized by that.

Images