JP3532116B2 - Information sorting method and system, recording medium - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an information selecting technique capable of selecting, from a wide variety of information, only information belonging to a user's taste range. In this specification, the expression “selection of information” is used as a concept including information retrieval.

[0002]

2. Description of the Related Art As a method of selecting information based on the idea that information of interest to people who have similar tastes to oneself is also information of interest to oneself, for example, United States Patent The collaborative filtering method disclosed in 4996642 is known.

In this collaborative filtering method, a person who has previously provided information used for preference determination (hereinafter referred to as "user") is asked to evaluate the information, and based on the evaluation result, the users The similarity (correlation value) of taste is calculated. Then, a user who requests information presentation or a user who presents selected information (hereinafter,
"Information presentation target person"), other users who have similar tastes, that is, a high correlation value of the evaluation result, are specified, and the information presentation target person is determined from the evaluation result by the specified user and the correlation value obtained earlier. The method of estimating the evaluation of the non-evaluated information of, and presenting the estimated highly evaluated information to the information presentation target person is adopted.

[0004]

Collaborative filtering is based on how similar the preference of one user is to the preference of another user, and the concept of preference range is incorporated. Has not been done. Therefore, when a wide variety of information is mixed, the overall taste is similar to the information presentation target person to some extent, but when viewed from a certain specific genre, the taste may not be so similar. In this case, the filtering did not work. In other words, when the information presentation target person has a wide variety of tastes, the correlation value based on the evaluation result of the portion where the tastes match is used in the same manner for information whose tastes do not match in the process of estimating the evaluation result. As a result, the latter information is also treated as a match.

[0005] For example, in the case where the information presenting person is interested in skiing and tennis and gives a high evaluation to the information about them, but is not interested in soccer, in the same manner as the information presenting person, skiing is performed. It is assumed that a plurality of users who give a high evaluation to tennis are identified as having similar tastes to the information presentation target person. It is quite conceivable that soccer information is included in the information that each of the identified users has highly evaluated. In this case, if it is attempted to estimate the evaluation of the information presenting person from the correlation value with the user who is similar to the other tastes of the information presenting person and the evaluation result of those users, the information about the soccer is also ski or tennis. The same high estimation result can be obtained as the above information. However, since the information presentation target person is not interested in soccer information, even if it is presented to the information presentation target person, it means that it does not match the preference of the information presentation target person.

In order to prevent such a situation, it is possible to classify the information to be presented in advance for each genre and perform collaborative filtering only within that genre. However, not all users evaluate information from the same viewpoint, and thus the accuracy of information selection may be significantly reduced depending on the classification standard. For example, one user may give a high evaluation to a product of a specific manufacturer, while another user may give a high evaluation to information on a function or a related product regardless of the manufacturer. In this way, it is not possible to determine from which viewpoint the genre classification is appropriate.

[0007] Therefore, the present invention is an improved information selection that can improve the accuracy of information selection even when a wide variety of presentation target information is mixed and the evaluation criteria of a plurality of users are indefinite. The task is to provide a method.
Another object of the present invention is to provide an information selection system to which the above information selection method is applied. Another object of the present invention is to provide a recording medium suitable for realizing the information selection method using a computer device.

[0008]

The information selection method of the present invention which solves the above-mentioned problems is one of the users, which holds a correlation value representing the similarity of tastes among a plurality of users for each user. In a method of selecting presentation target information that matches the preference of another user whose correlation value with the information presentation target is relatively high, some attributes are given to each of the plurality of presentation target information. Then, while updating the correlation value from time to time based on the evaluation result of each user for this attribute, among the presentation target information that matches the preference of the other user specified based on the updated correlation value. It is characterized in that the presentation target information to which the same attribute as the attribute that the information presentation target person relatively evaluates is given is selected. As a result, information that reliably reflects the taste is presented to the information presentation target person.

This information selection method can be implemented by, for example, a computer reading a predetermined program code and executing the following processing. This program code is recorded on a computer-readable recording medium. (1) A number of attributes are given to each of a plurality of pieces of presentation target information, and an information evaluation degree indicating the height of evaluation of the presentation pair information by each user and the evaluation of the user for each attribute A process of holding the attribute evaluation level indicating the height for all users, and updating the attribute evaluation level of the user for the attribute assigned to the relatively highly evaluated presentation target information to a higher level, ( 2) A process of identifying an attribute having a relatively high evaluation of the information presentation target person based on the updated attribute evaluation degree of the information presentation target person, (3) Relative to the same attribute as the identified attribute Process for identifying all other users who are highly evaluated, and updating the correlation value between the identified user and the information presentation target person to a higher value, (4) Correlation value after update Based on the degree of information evaluation of users who exceed a predetermined value The information presentation target person estimates the information evaluation degree for the presentation target information that has not yet been evaluated, and the information presentation target person relatively evaluates the information evaluation target person among the presentation target information corresponding to the estimated information evaluation degree. A process of presenting information including the same attribute as the attribute to the information presentation target person.
This processing is performed by the above-mentioned collaborative filtering, for example.

In the information selection system of the present invention which solves the above-mentioned other problems, some attributes are added to the information to be presented, and the information evaluation degree showing the height of evaluation of the information by the user. An information management unit that manages, for each user, an attribute evaluation degree that represents a height of evaluation of each of the attributes by the user, and a correlation value that represents a similarity degree of preference with other users, All other users who gave a relatively high evaluation to the same attribute as the one that the information presentation target person, who is one of the users, evaluated relatively high, based on each attribute evaluation degree A correlation value management unit that updates the correlation value between the identified user and the information presentation target person, and presents the information based on the information evaluation degree of the user whose updated correlation value exceeds a predetermined value. Subjects to be presented that have not been evaluated by the subject The information evaluation target for the information is estimated, and only the information including the same attribute as the information evaluation target person highly evaluates from the information to be presented corresponding to the estimated information evaluation level is estimated. And information presenting means for presenting the information to a person.

[0011] Preferably, the information management means is configured to update the attribute evaluation degree for the attribute assigned to each piece of presentation object information at any time according to the value of the information evaluation degree for the presentation object information. .

In a more specific embodiment, the information to be presented is product information of a virtual shopping mall formed by a predetermined computer system, and the information management means stores a plurality of product information for various product information. It is configured to update the correlation value of each user based on the history of the degree of interest indicated by the user, and the information presenting means extracts a part of the handled product information based on the updated correlation value. It is configured to be presented to the information presentation target person as recommended product information.

[0013]

BEST MODE FOR CARRYING OUT THE INVENTION Next, an embodiment of an information selection system to which the present invention is applied will be described. FIG. 1 is a configuration diagram of an information selection system 10 according to this embodiment. The information selection system 10 is realized by a computer device (or system) having an input device such as a keyboard, an output device such as a display, a communication mechanism, and a storage device, and the CPU of the computer device executes a predetermined program. Functional blocks of the input / output control unit 11, the information management unit 12, the correlation value management unit 13, and the information presentation unit 14, which are formed by reading and executing the code, and the information database DB and attributes built in the storage device. File F1, information evaluation file F2, attribute evaluation file F3,
The correlation value file F4 is provided.

The presentation target information database DB is adapted to store presentation target information (including evaluated information / not evaluated information) to be presented to the user.

The attribute file F1 is a file in which the correspondence between the identifier of the presentation target information in the presentation target information database DB and one or more attributes corresponding to the presentation target information is recorded. If the information to be presented is information about a product, the attribute is a product genre, its brand name or manufacturer, function, price, etc., and can be arbitrarily set according to the type of product, purpose, market characteristics, etc. it can.

The information evaluation file F2 is a file in which information evaluation data, which is an evaluation result of the information to be presented, is accumulated for each user (linked to the user identification data). This information evaluation data is an evaluation given by the user according to his / her preference for each piece of presentation target information, for example, 5
Numerical information such as “5 points”, “4 points”, ... If a user gives the camera 5 points and the television 4 points, the camera
Is stored in the information evaluation file F2 as "5 points" and the information evaluation data of "TV" is "4 points". Hereinafter, the evaluation result may be referred to as an “evaluation point” as represented by the above-mentioned numerical information.

The attribute evaluation file F3 is a file in which attribute evaluation data, which is an evaluation result for each attribute, is accumulated for each user. The evaluation result is the same as that used for evaluating the selection target information. The attribute evaluation data is updated at any time according to the value of the information evaluation data of the user, that is, the degree of interest in the information to be presented. When one attribute is given by a plurality of presentation target information, the evaluation result of the user for each of the plurality of presentation target information is reflected in the evaluation result of the attribute. For example, if the user has a strong interest in the high-class “camera” of company X, that is, gives a high evaluation, the attributes of this “camera” are “maker: company X” and “function: high function”. , A predetermined score is added to the attribute evaluation data “price: expensive” at any time. Similarly, when the user gives a high evaluation to the high-performance “TV” of company Y, “TV:” is the attribute “maker: Y”.
A predetermined score is added to the attribute evaluation data of "company" and "function: high function". At this time, the evaluation score of the attribute “function: high-performance” is reflected twice by the evaluations of both the “camera of X company” and the “TV of Y company”, and as a result, the score is added twice. It is a highly evaluated attribute.

The correlation value file F4 is a file for storing a correlation value indicating the similarity of preference among users for each user.

Next, the functional blocks 11 to 14 will be described. The input / output control unit 11 receives an information presentation request and other information input from an information presentation target person who is one of the users through an input device or a communication mechanism, and
Required information is presented to the information presentation target person.

The information management unit 12 manages the presentation target information and various kinds of information regarding the taste of each user, stores and reads out the information in the presentation target information database DB, and sets attributes for each presentation target information. , Various files F1 to F of the above-mentioned information evaluation data and attribute evaluation data
3 is composed of one or a plurality of modules that perform processing such as recording, updating, and reading.

The correlation value management unit 13 calculates the correlation value for each user using a known calculation formula (for example, Pearson product moment correlation coefficient calculation formula), and stores the calculation result in the correlation value file F4. It is a module for storing and updating the correlation value of each user as needed based on the information evaluation data and attribute evaluation data of each user. Specifically, the other users B to D who have given a high evaluation to the same attribute that a certain user A highly evaluates are identified based on the respective attribute evaluation degrees, and the identified users A process of updating at any time so that the correlation value between B to D and the user A is increased is performed.

The information presenting unit 14 applies to the presentation target information which has not been evaluated by the information presentation target person based on the information evaluation degree of the user whose latest correlation value with the information presentation target person exceeds a predetermined value. The information evaluation degree is estimated, and only the information including the same attribute as the information evaluation target person's relatively high evaluation is presented to the information presentation target person from the presentation target information corresponding to the estimated information evaluation degree. It is a module that performs processing for doing so.

The functional blocks (modules constituting the functional blocks) 11 to 14 and the presentation target information database D
B and the program code for forming each of the files F1 to F4 on the computer device are usually recorded on a fixed recording medium such as a hard disk or a semiconductor memory, and read and executed by the CPU of the computer device at any time. It is supposed to be done. However, since the present invention can be implemented if the above-mentioned functional blocks 11 to 14 are formed in the computer device during operation, the recording form thereof may be arbitrary. For example, it is recorded on a portable recording medium such as a CD-ROM, a DVD, an FD or the like which can be separated from a computer device, or a program server connected to a local network, read when used, and installed on the fixed recording medium. It may be provided for execution at any time. The above functional blocks 11 to 14
The above may be realized not only by the above-mentioned program code alone, but also appropriately realized by cooperation with the operating system installed in the computer device.

Next, the operation of the information selection system 10 will be described with reference to FIG. Here, it is assumed that the attribute information file F1 has already been created. Further, it is assumed that the information evaluation data of each of the plurality of users (including the information presentation target person) is accumulated in the information evaluation file F2. Further, when the information presentation target person requests the information presentation, the presentation target information is selected and presented. However, this premise is for convenience of explanation, and such a premise is not always indispensable.

In FIG. 2, the information management unit 12 refers to the information evaluation file F2 and the attribute evaluation file F3, and the evaluation by each user is relatively high (for example, the maximum evaluation point or a predetermined evaluation point is exceeded). ) The presentation target information and the attribute corresponding thereto are specified based on the association information in the attribute information file F1 (step S101).
Then, the attribute evaluation data for the identified attribute is updated. Specifically, the evaluation points are added (step S10).
2). This is repeated for all applicable users (step S103: No).

When the attribute evaluation data of each user is updated (step S103: Yes), the correlation value management unit 13
The correlation value of the preference between the users is calculated, and the correlation value of each user in the correlation value file F4 is updated based on the calculation result (step S104). Upon the input of the information presentation request from the information presentation target person (step S105: Yes), the latest attribute evaluation data of the information presentation target person in the attribute evaluation file F3 is referred to, and the attribute having a relatively high evaluation among them. The evaluation data and its attributes are specified (step S1
06), another user who is similarly highly evaluated for the specified attribute is specified (step S107). This processing is performed by the information management unit 12. Whether or not the attribute has a high evaluation is determined by, for example, whether or not an evaluation score of a certain value or more is attached, or whether the ratio of the entire attribute to the evaluation score is a certain value or more.

Thereafter, the correlation value management unit 13 updates the correlation value file F4 by adding α points to the identified correlation value of each user (step S108). If the α point is large, only the presentation target information with a high evaluation score for the same attribute will be extracted. Conversely, if the α point is small, the evaluation score for the attribute will not be reflected so much. The value of α point is a tuning item in the system.

The information presenting section 14 refers to the correlation value file F4 and the information evaluation file F2, and presents information by using the information evaluation data of the user whose updated correlation value exceeds a predetermined value and the correlation value. The evaluation of the presentation target information that the target person has not yet evaluated is estimated (step S109), and some pieces of presentation target information with high estimated evaluation are specified (step S110). After that, these presentation target information,
It narrows down to the information corresponding to the attribute highly evaluated by the information presentation target person (step S111), and presents this presentation target information to the information presentation target person through the input / output control unit 11 (step S112).

By limiting the range of presentation target information in this way, even when various types of presentation target information are mixed, information selection and reflection that reliably reflect the preferences of the information presentation target person can be performed. Can be presented. Therefore, by applying the method of the present embodiment to the collaborative filtering, the accuracy thereof can be significantly improved as compared with the conventional method.

[0030]

EXAMPLE The information selection system 10 of the present embodiment is capable of accurately narrowing down and presenting information that matches each taste to a plurality of users in a situation where various types of presentation target information are mixed. . Here, as an example, a case where the invention is applied to a virtual shopping mall (shopping mall) realized on the Internet will be described.

In this case, the information selection system 10 has one web server (a plurality of servers when performing distributed processing) connected to the Internet, the above-mentioned presentation target information database DB, each file F1 to F4, and each function. The blocks 11 to 14 will be formed. Information on various products handled in the shopping mall is stored in the presentation target information database DB. In each of the files F1 to F4,
Information evaluation data, attribute evaluation data, and correlation values for registered users are stored.

The information management unit 12 is configured to automatically update the correlation value of each user based on the history of the degree of interest indicated by a plurality of users with respect to various product information handled in the shopping mall. More specifically, the degree of interest of the user, that is, the degree of evaluation of the information, is quantified based on the reference history of each user and the purchase history of the product for the products offered in each store of the shopping mall. The above-described information evaluation data and attribute evaluation data are generated and updated, and the evaluation value based on these is updated. The information presenting unit 14 presents the handling product information selected based on the updated correlation value to the information presentation target person as recommended product information, for example.

In the actual operation of the shopping mall, in addition to the information selection system 10 described above, an operation management server 20 for exchanging information with a store server (agent) constructing the mall, a computer provided in a credit company A payment server 30 and the like that cooperate with the system to perform payment processing when a user purchases a product will be provided.

By applying the information selection technique of the present invention to a shopping mall as in this embodiment, information regarding the tastes of a plurality of users with respect to the products handled is automatically collected, and based on the collected information. Since recommended product information narrowed down to the range of tastes can be presented to each user, the user's awareness of using the shopping mall can be increased.

[0035]

As is apparent from the above description, according to the present invention, some attributes are added to presentation target information,
Since the information is selected after reflecting the evaluation result of the assigned attribute on the correlation value of the preference between users, various types of presentation target information are mixed and the evaluation criteria of multiple users are undefined. Even in such a case, there is a unique effect that the accuracy of information selection can be improved.

[Brief description of drawings]

FIG. 1 is a functional block diagram showing an embodiment of an information selection system of the present invention.

FIG. 2 is a procedure explanatory view showing an example of an operation (information selection method) of the information selection system of the present embodiment.

FIG. 3 is an explanatory view showing an embodiment of the present invention.

[Explanation of symbols]

10 Information sorting system 11 I / O controller 12 Information Management Department 13 Correlation value management unit 14 Information presentation section F1 attribute information file F2 information evaluation file F3 attribute evaluation file F4 correlation value file 20 Operation Management Server 30 payment server

Continuation of the front page (56) References JP 10-340270 (JP, A) JP 10-240749 (JP, A) JP 8-320879 (JP, A) Ariyoshi, using multivariate analysis Social Information Filtering, Research Report of Information Processing Society of Japan (98-FI-51), September 18, 1998, Vol. 98, No. 81, p. 17-24 (58) Fields investigated (Int.Cl. ⁷ , DB name) G06F 17/30 JISST file (JOIS)

Claims (10)

(57) [Claims] 1. A computer provided with a correlation value file for storing a correlation value indicating a degree of preference similarity among a plurality of users for each user.
Computer devices are each subject to user evaluation
Multiple presentation target information with some attributes set in advance
From the information, a method of selecting the presentation target information that the correlation value with the information presentation target person who is one of the users is relatively high, which is suitable for the preference of another user, in the correlation value file Evaluation of the stored correlation value of each user with respect to the attribute of the presentation target information by the user
Information to be presented that is updated from time to time based on the result and that matches the preference of the other user based on the updated correlation value
And selecting the presentation target information to which the same attributes as the attributes that the information presentation target person relatively highly evaluates are assigned from the specified presentation target information. . 2. The information selection method according to claim 1, wherein the computer device cumulatively enhances the evaluation result of the user with respect to the attribute added to the presentation target information highly evaluated by the user. . 3. The computer device, when one attribute is provided by a plurality of pieces of presentation target information, reflects the evaluation result of the user for each of the plurality of pieces of presentation target information in the evaluation result of the attribute. The information selection method according to claim 1, wherein: 4. A computer provided with a correlation value file for storing a correlation value indicating a degree of preference similarity among a plurality of users for each user.
Computer to execute the evaluation pairs performed by each user.
Multiple proposals with some elephant attributes preset
From shows target information, a method for selecting the presentation target information matches the preference of the correlation value is relatively high other users between the information target person is one of the user, individual The information evaluation degree indicating the evaluation height of the presentation versus information by the user and the attribute evaluation degree indicating the evaluation height of the user for each attribute are held for all users, and are relatively highly evaluated. The process of updating the attribute evaluation degree of the user concerned about the attribute given to the presentation target information to a higher degree, the evaluation of the information presentation target person is relatively based on the attribute evaluation degree of the information presentation target person after the update. To identify attributes that are relatively high, identify all other users who give a relatively high evaluation to the same attributes as the identified attributes, and identify the identified users and the information presentation target person. Update the correlation value between Based on the information evaluation degree of the user whose updated correlation value exceeds a predetermined value, the information evaluation degree for the presentation target information that has not yet been evaluated by the information presentation target person is estimated, and the estimated information evaluation degree is calculated. An information selection method, which includes a process of presenting, to the information presentation target person, information including the same attribute as the attribute that the information presentation target person relatively highly evaluates from the corresponding presentation target information. . 5. The information selecting method according to claim 4, wherein the information that has not been evaluated yet is selected by collaborative filtering. 6. Each is subject to evaluation by the user
Information evaluation level indicating the evaluation level for each of the plurality of pieces of presentation target information in which some attributes are preset, and attribute evaluation level indicating the evaluation level of each of the attributes by the user , And the information management means for managing the correlation value indicating the degree of similarity of preference with other users for each user, and the information presentation target person who is one of the users highly evaluates it. All other users who have given a relatively high evaluation to the same attribute as the attribute are specified based on their respective attribute evaluation degrees, and the correlation value between the specified user and the information presentation target person And a correlation value management means for updating, and estimates the information evaluation degree for the presentation target information not yet evaluated by the information presentation target person based on the information evaluation degree of the user whose updated correlation value exceeds a predetermined value. , Information to be presented corresponding to the estimated information rating And an information presenting means for presenting to the information presenting person only information including the same attribute as the information presenting person's relatively high evaluation. 7. The information management means is configured to update an attribute evaluation degree for an attribute assigned to each piece of presentation target information at any time according to a value of the information evaluation degree for the presentation target information. The information selection system according to claim 6, wherein: 8. The information to be presented is product information of a virtual shopping mall formed by a predetermined computer system, and the information management means has a degree of interest indicated by a plurality of users with respect to various product information. Is configured to update the correlation value of each user based on the history of the information presenting means, the information presenting means presents a part of the handled product information as recommended product information based on the updated correlation value. The information selection system according to claim 6 or 7, which is presented to the target person. 9. The information management means is configured to quantify the degree of interest of each user based on the reference history and purchase history of each user with respect to the products handled in the virtual shopping mall. The information selection system according to claim 8. 10. A correlation value file for storing, for each user, a correlation value representing a similarity of preference among a plurality of users, which is provided .
Each attribute has some attributes that are evaluated by the user.
A function of selecting presentation target information that matches the preference of another user who has a relatively high correlation value with the information presentation target person who is one of the users from the plurality of presentation target information set for In a computer device equipped with, for all users, an information evaluation degree indicating the evaluation level of the presentation-to-information by an individual user and an attribute evaluation degree indicating the evaluation level of the user for each attribute holds, the process of updating the increase the user's attribute evaluation degree of the attributes that have been granted to the relative acclaimed presentation target information, based on the information the target person attribute evaluation of the updated Process to identify attributes that are relatively highly evaluated by the information presentation target person, and to identify and identify all other users who are relatively highly rated to the same attributes as the identified attributes. Users and the above information A process of updating the correlation value to a higher value with the presentation target person, a presentation target that has not been evaluated by the information presentation target person based on the information evaluation degree of the user whose updated correlation value exceeds a predetermined value The information evaluation target for the information is estimated by estimating the information evaluation degree for the information, and the information including the same attribute as the information evaluation target person's relatively high evaluation among the presentation target information corresponding to the estimated information evaluation degree. A computer-readable recording medium on which is recorded a program code for executing the processing presented to.