JPH11282874A - Information filtering method and device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain a highly accurate filtered result even when a profile is a polysemous work. SOLUTION: When a profile is inputted by as user (a step S1), filtering based on the profile is performed and its result is displayed to the user by an information filtering part (a step S5). The filtered result is evaluated by the user (a step S6), and when information resources in which no interest is shown by the user exist, another user having interest in the information resources is searched by the information filtering part (a step S8). Next, correlation between such other user and the user as a requester for filtering is searched. When no correlation exists between both users, the filtered result by the other user's profile is excluded from the result outputted in the step 5 and is outputted as a new filtered result (a step S12).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an information filtering method and apparatus for filtering information resources, and more particularly, to an information filtering method and apparatus for combining user profiles in multiple stages according to user preferences.

[0002]

2. Description of the Related Art There are the following documents that show conventional information filtering techniques. [1] Nicholas J. Belkin, W. Bruce Croft, Information
Filtering and Information retrieval: Two Sides of t
he Same Coin ?, Communication of the ACM, 35 (12), pp29
-38,1992

[2] Upendra Shardanand, Pattie Maes, S
ocial Information Filtering: Algorithms for Automat
ing "Word of Mouth", CHI'95 Conference Proceedings,
1995, pp210-217

[3] Chris Beckley, Gerard Salton, Ja
mes Allan, The Effect of AddingRelevance Informati
on in a Relevance Feedback Environment, SIGIR'94 Pr
oceedings, 1994, pp292-300

[0005] With the development of computer networks, information filtering technology has been developed that obtains only the information desired by the user for information that changes every day, such as e-mail and net news. In information filtering, related feedback (Relevance Feed)
In general, a system that modifies a user's profile by using a back algorithm or a learning algorithm to perform filtering.

[0006] Here, a profile has almost the same meaning as a search sentence in information search or a question sentence, but here, a search sentence expressing a user's interest in search for a certain length of time. It is. Such filtering is performed by pattern matching of a profile expressing the text content and information desired by the user, and this is called content-based filtering (this is described in, for example, the above-mentioned document [1]).

[0007] The procedure of information filtering is generally performed by the following method. 1. The user enters an initial profile. 2. The information filtering device compares the user's profile with the delivered text, performs filtering, and displays the filtering result to the user. 3. The user inputs an evaluation for the filtering result. Repeat steps 4.2 to 4.

[0008]

However, the above-mentioned conventional information filtering apparatus has the following problems. First, in the information filtering procedure,
The user inputs whether the user is interested or not interested in the filtered article. At a stage where the number of repetitions of steps 2 to 4 in the filtering procedure is small, the profile is sufficiently modified. There was a problem that the accuracy of the filtering did not improve much.

[0009] A second problem is that profiles generally do not contain enough information to filter information resources. Therefore, the specific information (a certain base word or a certain feature vector) in the profile has a different importance for each data field. Therefore, if you try to filter multi-disciplinary data with the same profile,
There is a base word of a profile in which the importance is high in one specific field but low in another field, and the accuracy of filtering is reduced as a whole. Note that the base word is
As described in the specification of Japanese Patent Application No. 9-157909 previously filed by the inventor, the present invention constitutes a vector axis in a vector space model. Further, the vector space model will be described in detail in the embodiment section.

[0010] For example, in the vector space model, polysemy words are expressed as the same feature vector, which is one factor that deteriorates search accuracy. As an example, it is assumed that a database engineer of a computer designates a profile of “model”, and a text of “model” related to fashion is obtained as a filtering result. Here, the word "model" is polysemous because it has different meanings depending on the context. If you try to refine the profile by using related feedback, the base word "model" becomes less important, and in the next filtering process, it is necessary to filter the text containing the base word "model" in the database. I can no longer do it.

Further, in the case of filtering, it is often the case that a database which is not classified by field is targeted, and it is difficult to improve the accuracy of filtering such a database by polysemy.

[0012] In view of the above, there has been a demand for an information filtering method and apparatus capable of performing high-accuracy filtering even if the profile is a polysemy.

[0013]

The present invention employs the following structure to solve the above-mentioned problems. <Structure of Claim 1> Information resources conforming to a user's profile are collected, other users using the same profile in different meanings are extracted, and a correlation between another user and the user is obtained. If there is no information resource, the information filtering method is an information filtering method in which information resources of interest to other users are excluded from the collected information resources and a filtering result for the user is obtained.

<Explanation of Claim 1> The fact that the same profile is used in different meanings means that, for example, a "model" related to a computer database differs from a "model" related to fashion in a profile of polysemy. It refers to the case where it is used in a meaning. The correlation between the users is determined, for example, based on the occupation of the user as to whether or not there is a correlation. If the condition is predetermined, the correlation is determined based on something other than the occupation. It may be.

That is, according to the first aspect of the present invention, a filtering result of a user who uses the same profile in a different sense from a filtering result of a certain profile and has no correlation with the user is excluded. This is the result of the filtering.

According to the first aspect of the present invention, information resources unrelated to the user can be deleted, and as a result, the accuracy of filtering can be improved.

<Structure of Claim 2> Information resources conforming to the user's profile are collected and output to the user. As a result, among the output information resources, information resources that the user is not interested in are output. Is extracted, and other users who use the same profile as the user are extracted. Then, a correlation between the user and the other user is obtained. If there is no correlation, the information resource which the other user is interested in is obtained. Is excluded from the collected information resources to obtain a new filtering result for the user.

<Explanation of Claim 2> According to the invention of claim 2, as a method of extracting other users using the same profile in different meanings according to the invention of claim 1, first, a filtering result based on an initial profile is output. Then, a user who is interested in an information resource such that the user is not interested in the result and uses the same profile is extracted as such another user.

According to such a configuration, a highly accurate search result can be obtained without the user repeating the operation such as modifying the profile.

<Structure of claim 3> In the information filtering method according to claim 1 or 2, the correlation between the user and another user is determined by performing social filtering. It is.

<Explanation of Claim 3> The invention of claim 3 is the invention of claim 1 or 2, wherein the correlation between users is determined by using a social filtering function. Thereby, the same effect as the first or second aspect of the invention can be obtained.

<Configuration of Claim 4> A user information storage unit for storing attribute information of a plurality of users, and personal information management for managing information indicating a correlation with a plurality of user attributes stored in the user information storage unit. And a resource including a profile input by a specific user as a filtering result, and for the filtering result,
If there is an information resource that responds that the specific user is not interested, the other users who are interested in the information resource and use the same profile as the specific user are extracted. The personal information management unit inquires whether there is a correlation with the personal information management unit. If there is a response from the personal information management unit indicating that there is no correlation, the information resource including the profile of the specific user indicates that another user is interested. And an information filtering unit that excludes the information resource that has been set as a new filtering result for a specific user.

<Explanation of Claim 4> The attribute information of the user is, for example, an attribute such as a computer engineer or a person who works in the fashion industry. It may be something. In addition, the information indicating the correlation with respect to the attribute of the user is, for example, information such as “there is no strong correlation” between the “worker working in the fashion industry” and the “worker working in the computer industry”. Any information may be used as long as the information indicates the relationship.

The information filtering unit first outputs a filtering result based on a profile input by a specific user. Then, the specific user inputs whether or not he or she is interested in this result. Thus, the information filtering unit extracts another user who is interested and uses the same profile for the information resource that the specific user has input that he or she is not interested in. Next, the information filtering unit inquires of the personal information management unit about the correlation between another user and the specific user.

Thus, the personal information management unit refers to the user attribute information in the user information storage unit, checks whether there is a correlation between both users, and notifies the result to the information filtering unit. If the result is "no correlation", the information filtering unit excludes information resources that are of interest to other users from the information resources including the profile of the specific user and outputs the result as a new filtering result I do.

According to the fourth aspect of the present invention, a highly accurate search result can be obtained without repeating the operation such as modifying the profile by the user.

<Structure of Claim 5> A user evaluation information storage unit for storing information evaluated by an arbitrary user with respect to a specific information resource, a user requesting information filtering, and a plurality of other users A social filtering unit that evaluates the same information resource and extracts a user having the lowest similarity to the requester user from among a plurality of other users based on the evaluation value; An information resource including the input profile is output as a filtering result, and if there is an information resource that responds that the specific user is not interested in the filtering result, the user is interested in the information resource and the specific user Extract other users using the same profile as
Among other users, the social filtering unit inquires who is the user with the lowest similarity to the specific user, and the information resource that the user with the lowest similarity responded by the social filtering unit is interested in, An information filtering device comprising: an information filtering unit that excludes information resources including a profile of a specific user and sets a new filtering result.

<Explanation of Claim 5> According to the invention of claim 5, another user who is interested in an information resource responding that the specific user is not interested and uses the same profile as the specific user. Among the users, a user having the lowest similarity to a specific user is determined using a social filtering function. With such a configuration, similarly to the invention of claim 4, there is an effect that a highly accurate search result can be obtained even if the user does not repeat an operation such as correcting a profile.

[0029]

Embodiments of the present invention will be described below in detail with reference to the drawings. Embodiment 1 <Configuration> FIG. 1 is a flowchart showing an information filtering method according to Embodiment 1 of the present invention. Prior to the description, an information filtering apparatus for implementing the information filtering method will be described.

FIG. 2 is a block diagram showing a first embodiment of the information filtering apparatus of the present invention. The apparatus shown in the figure comprises a user information storage unit 1, a personal information management unit 2, an interface management unit 3, an information filtering unit 4, a message processing unit 5, and an information providing unit 6, and these components are realized on a computer.

The user information storage unit 1 is provided in a storage device such as a hard disk device of a computer, and stores attribute information of a plurality of users. Here, the attribute of the user is, for example, the occupation of the user.
As an example, one user is “a person who works for a fashion-related company” and another user is information such as “a computer engineer”.

The personal information management unit 2 includes a user information storage unit 1
Has a function of managing information indicating a correlation between attributes of a plurality of users stored in the. For example, the relationship between “workers in the fashion industry” and “workers in the computer industry” has information such as “there is little correlation”.

The interface management unit 3 (3a, 3b, 3
c) is provided for each of the plurality of users 20a, 20b, and 20c, receives an input of each user, sends information input by the user to the information filtering unit 4, and filters the information sent from the information filtering unit 4. It has a function of displaying the result to the corresponding user.

The information filtering unit 4 (4a, 4b, 4
c) is provided for each user corresponding to the interface management unit 3, and the profile of the corresponding user and the information source 1 are provided.
It has a function of comparing with 0 to perform filtering based on pattern matching and a function of modifying a profile based on user feedback. Further, it has a function of converting a profile sentence input by a user into base words and generating a feature vector.

That is, the information filtering unit 4 is a specific user (a user corresponding to the own information filtering unit 4).
Manages the input profile, performs filtering from the information source 10 based on the profile,
The profile is modified depending on whether or not the specific user is interested in outputting the filtering result, and a user who is interested in the information resource instructed that the specific user is not interested is searched for. The personal information management unit 2 is inquired as to whether there is a correlation with a specific user, and if a response indicating that there is no correlation is received from the personal information management unit 2, the filtering result of the specific user's profile is It has a function to delete the result filtered by the profile and make it a filtering result for a specific user.

Although only three users 20, the interface management unit 3 and the information filtering unit 4 are shown in the drawing, there are actually four or more users and these users Assume that an interface management unit 3 and an information filtering unit 4 are provided correspondingly.

The message processing unit 5 has a function of providing a communication place for the information filtering unit 4, the information providing unit 6, and the personal information managing unit 2. Further, the information providing unit 6 has a function of transmitting an identifier (Identifier (id)) of the information resource in the information source 10 to the message processing unit 5. Here, the information resource indicates one unit of information included in the information source 10, and the identifier is information for identifying each information resource. In this specific example, this identifier is called a document identifier.

The information source 10 includes a plurality of information resources as described above, and includes data of documents and images.
The filtering device is connected to the information source 10 through a computer network.

Each of the above-mentioned interface management unit 3 to information providing unit 6 is composed of a program for realizing each function on a computer, and a processor and a memory for executing these programs.

<Operation> In the specific example 1, it is assumed that a user (this is referred to as a user 20a) wants to filter articles of a magazine related to a "model" in a book on a computer database. If the user 20a registers a profile in the filtering device, after the registration, the information filtering device deletes an article of a magazine related to the "model" every time an article of the magazine enters a certain information resource. Deliver to the user.

The operation of the first embodiment will be described below with reference to FIG.

[Step S1] The information filtering device is activated by a user activation command. At this time, the information filtering device waits for the user 20 to input data. Now, the user 20a has a profile
It is assumed that the word "model" is input via the interface management unit 3a. In addition, as the personal information of the user 20a, the user inputs that he is a "computer engineer".

[Step S2] Interface management unit 3
a is the profile input in step S1,
The information is sent to the information filtering unit 4a. The information filtering unit 4a creates a feature vector from the profile of the user 20a.

Here, a vector space model for filtering by the information filtering unit 4a and a related feedback (Relevance Feedback) will be described. The vector space model and the related feedback are used as one method for performing information filtering. In addition to this, the profile is refined using other methods, for example, a probability model in information retrieval, and machine learning. It is also possible to do so.

The vector space model is a search method that takes a word in a text as a vector (for example, this is described in reference [3]). The text is divided into words, and the words are assigned importance levels, which are used as feature vectors. Here, in general, on a distributed network, information resources include image data, moving image data, compressed files, and the like, but in this specific example, information resources are limited to text. Also, a natural language sentence is used for the question sentence, and the question sentence is converted into a feature vector, similarly to a text (or document) word. The search result ranks the similarity between the text and the question sentence. Hereafter, the word is called a base word

A text vector is represented by the following equation. FIG. 3 is an explanatory diagram of the arithmetic expression of the vector space model. In FIG. 3, (1) is an expression indicating a vector DW of a certain text. In (1), dw ₁ , dw ₂ ,..., Dw
_t is the importance of the base word of the text.

Similarly, the vector Q of the question sentence is expressed by equation (2). Here, q ₁ , q ₂ ,..., Q _t are the degrees of importance of the base words of the question sentence.

[0048] Further, the base word importance (weight vector word T _k of the document D _i) is given by equation (3). here,
The meaning of each symbol is as follows. · W _ik: document D _i base language T _k of a vector of weighting · tf _ik of: document D _i base language T _k of the number of occurrences · N of: The total number · n _k of the collected documents: in the collection documents And the number of documents containing the base word T _k

The similarity Sim is given by the following equation (4). Here, there is a threshold such as θ in (5). The search results are displayed in the order of the degree of similarity among those satisfying the expression (3).

A method in which a user evaluates a retrieved document and feeds a vector of the document that the user is related to the question sentence and refines the retrieval formula is referred to as a related feedback.
For example, it is shown in reference [3]).

Various methods have been proposed for the related feedback. In general, for example, equation (6) as shown in Ref. [3] is used. Here, rel_docs is a feature vector of a document of interest to the user among the retrieved documents. Further, nonrel_docs is a feature vector of a document that the user is not interested in among the retrieved documents. In general, α, β, γ have the value 8,
16, 4 are used respectively.

[Step S 3] The information providing unit 6 sends the document identifier of the information resource sent from the information source 10 to the message processing unit 5.

[Step S4] Information filtering section 4
a, when the document identifier of the information resource enters the message processing unit 5, filters the result and sends the result to the interface management unit 3a. Here, the filtering is performed by the equations (1) to (5) of the vector space model.

[Step S5] Interface management unit 3
a displays the filtering result to the user.

FIG. 4 is an explanatory diagram of the operation of the information filtering unit 4a. The state shown in FIG.
9 shows the flow of processing from S5 to S5. That is, a document identifier, which is an identifier of an information resource, is input from the information source 10 to the information filtering unit 4a via the information providing unit 6. The information filtering unit 4a performs filtering on such a document identifier with reference to the profile of the user 20a, and outputs a filtering result.

[Step S6] The user 20a evaluates the filtering result. Here, the user 20a inputs to each information resource whether the information is interested or not.

[Step S7] Information filtering section 4
a modifies the profile based on the evaluation information of the user 20a in step S6. Here, the correction of the profile is performed by the expression (6) of the related feedback.

[Step S8] Information filtering section 4
a inquires of the message processing unit 5 whether or not there is a user who is interested in the article for which the user 20a has input that the user is not interested.

[Step S9] Assume that another information filtering unit 4 has responded to the message of step S8.
Here, assuming that the information filtering unit 4b has responded, the message is sent to the message processing unit 5.

[Step S10] The information filtering unit 4a inquires of the personal information management unit 2 about the correlation between users with respect to the information filtering unit 4b responded in step S9.

[Step S11] The personal information management unit 2
Answer the correlation between users. Here, the personal information management unit 2
Responds to the message in step S10 with the user 20a
And the user 20b send to the message processing unit 5 that "there is no strong correlation".

Here, a specific method for obtaining the correlation between users will be described. FIG. 5 is an explanatory diagram showing classification of users. As shown, the users are classified into “computer”, “fashion”, “economy”, etc., and the users related to “computer” are classified into users such as “human interface”, “database”, “computer network”, etc. Is done. For example, the user 20a here falls into the category of “database” in the category of “computer”.

In the personal information management unit 2, if the classification of the users is different at the highest level in the relationship as shown in FIG. 5, it is determined that there is no correlation between the users. In the case of the above example, since the user 20a belongs to the category of “computer” and the user 20b belongs to the category of “fashion”, it can be said that there is no correlation between the user 20a and the user 20b.

[Step S12] The information filtering unit 4a performs filtering using the corrected profile and the profile of another user 20b. Here, the information filtering unit 4a performs filtering except for information resources that are filtered by the profile of the information filtering unit 4b. At this time, the relationship among the user 20a, the interface management unit 3a, the information filtering units 4a and 4b, the information providing unit 6, and the information source 10 is as follows.

FIG. 6 is an explanatory diagram showing the order of filtering. As shown in the figure, the document identifier (information resource) of the information source 10 provided by the information providing unit 6
Filtering (exclusion filtering) by the profile of the information filtering unit 4b is performed, and then filtering is performed by the profile of the information filtering unit 4a, and the result is output to the user 20a via the interface management unit 3a.

[Step S13] The user 20a inputs whether to end the filtering.

[Step S14] In step S13, if the user 20a inputs the end of the filtering, the filtering device ends the filtering.

<Effects> As described above, according to the first embodiment, the profile of another user whose base word (or keyword) in the profile of the user is used as a profile having a different meaning is used. By searching and using the profile, information resources not related to the user can be deleted, and as a result, the accuracy of filtering can be improved.

<< Example 2 >> In Example 2, a correlation between users is determined using a filtering function called social filtering or colaborative filtering.

Here, the social filtering is the following filtering. That is, a user who is a requester of information filtering and a plurality of other users evaluate the same information resource, and based on the evaluation value, a user who is a requester among other plurality of users is evaluated. The user with the highest similarity is extracted, and for the information resource for which the user who is the requester of the information filtering has not been evaluated,
Using the information on how the user having the highest similarity has evaluated this information resource, the information resource that has not been evaluated is filtered.

The social filtering is a method of collecting recommendation information of a user and information of personal preference and filtering based on the information. Social filtering is performed on media, such as movie information and music information, whose contents are difficult to express on a computer. For example, sounds, photos, arts, videos, multimedia information, etc. are difficult to find attributes associated with the user's profile and perform content-based filtering. Therefore, social filtering is effective for such information.

Specifically, the following processing is performed in social filtering. 1. The user saves the input information and the evaluation information for each information resource (here, book, video, music, etc.). 2. The information filtering device calculates a correlation between the evaluation result of the newly registered user and the evaluation result of another user, and obtains an information resource (here,
Book, video, music, etc.). 3. When a new information resource is registered, 2. Is calculated, and the evaluation value of the user for the information resource is predicted. 4. An information resource with a high evaluation value is presented to the user. 5. Each time a new information resource is registered, 2. ~
4. Is repeated.

<Structure> FIG. 7 is a diagram showing the structure of the second embodiment. The illustrated device includes an interface management unit 3, an information filtering unit 4, a message processing unit 5, an information providing unit 6, a user evaluation information storage unit 7, and a social filtering unit 8.
Consists of Here, the interface management unit 3 to the information providing unit 6 are the same as those in the first embodiment.

The user evaluation information storage section 7 stores information on an arbitrary user evaluated for a certain information resource. The social filtering unit 8 has the following functions.

1. The user evaluation value is stored in the user evaluation information storage unit 7 for each information resource. 2. The above 1. Calculate the correlation between the user's evaluation value and the other user's evaluation value input in step (1), and predict the evaluation value of the information resource that the user has not seen yet. 3. An information resource with a high evaluation value is presented to the user. 4. Each time a new information resource is registered, the above 2.3.
repeat.

As described above, the social filtering unit 8 has a function of selecting another user's profile similar to a certain user's profile. Further, it has a function of finding a profile of another user that does not match the interest of a certain user.

<Operation> Also in the specific example 2, the user
Suppose you want to filter out articles in magazines related to "models" among magazines related to databases in the field of computers. If the user registers a profile in this filtering device, the information filtering device deletes an article of a magazine related to the “model” every time an article of the magazine enters the information providing unit 6 from the information source 10. Deliver to.

FIG. 8 is a flowchart showing the main part of the operation of the second embodiment. Here, it is assumed that the state is after step S10 in the first specific example.

[Step S1] Information filtering section 4
Sends, to the message processing unit 5, a message indicating whether there is any user who is interested in the article with respect to the article that the user 20 is not interested in.

[Step S2] The social filtering section 8 sends a reply to the message in step S1 to the message processing section 5 to search for a suitable information filtering section 4. Here, it is assumed that the social filtering unit 8 has found the information filtering units 4b to 4f. FIG.
Is an explanatory diagram of that state.

[Step S3] The social filtering unit 8 finds an information filtering unit 4 that does not match the interest of the user 20a. That is, the similarity with the information filtering units 4b to 4f is calculated by Expression (7) shown in FIG. 9, and the one with the lowest similarity is the user 20a.
To obtain an information filtering unit 4 that does not match the interest of the user. Here, it is assumed that the information filtering unit 4 that does not match the user's interest is the information filtering unit 4c.

Here, “how to find the information filtering unit 4 that does not match the interest of the user 20a” will be described. First, the similarity scores U _y of the user with other and scores U _x of a user, for example, can be calculated from equation (7) in FIG. For details, see [2]
It is also shown in

In the figure, r _xy indicates the similarity, and U _x and U _y with overlines indicate the average of the scores of the users and the average of the scores of the other users, respectively. Here, the score indicates the degree of interest given to a certain information resource by the user. This degree of interest is, for example, assigned a score with the numbers 1,2,3,4,5,6,7, where 1 indicates that the user is not interested in the information resource at all, and 7 indicates Indicates that the user is very interested in the information resource.

Such an expression is called a Pearson r algorithm. The r _xy used in the above equation is applied to all other users to determine the degree of interest between the users.

[Step S4] By step S3,
Filtering is performed using the found information filtering unit 4. Here, the information filtering unit 4a performs filtering except for information resources that are filtered by the profile of the information filtering unit 4c. The order of filtering at this time is as follows.

FIG. 10 is an explanatory diagram showing the order of filtering. As shown in the drawing, the document identifier (information resource) of the information source 10 provided by the information providing unit 6 is first subjected to filtering (exclusion filtering) by the profile of the information filtering unit 4c.
The information filtering unit 4a performs filtering based on the profile, and outputs a result of the filtering to the user 20a via the interface management unit 3a.

<Effects> As described above, according to the specific example 2, the social filtering function is used to search for a profile that uses the base word as a different meaning in the user's profile. , Information resources unrelated to the user can be deleted. As a result, there is an effect that the accuracy of the filtering can be improved even when the number of times of feedback for correcting the profile is small.

<< Usage Form >> The present invention is not limited to the above-mentioned specific examples, but can be applied to the following modified examples.

[0109] Instead of the information providing unit 6, the present invention can be applied to a connection to a normal information retrieval system.

Instead of the information providing unit 6, the present invention can be applied to a system connected to a system that receives articles such as texts and images from a plurality of information sources (for example, newspaper companies, publishers, and news agencies). .

In the first and second embodiments, the vector space model is used as the information filtering technique. However, a probability model used in information retrieval may be used.

In the first and second embodiments, the related feedback is used as a method for correcting the profile of the user. However, a machine learning method may be used.

In the specific examples 1 and 2, the information source 10 is one, but the invention can be applied to a plurality of information sources.

Instead of the Poisson r algorithm used in the second embodiment, another algorithm for determining the correlation between users may be used.

[Brief description of the drawings]

FIG. 1 is a flowchart of an information filtering method according to a first embodiment of the present invention.

FIG. 2 is a configuration diagram of an information filtering device according to a first embodiment of the present invention.

FIG. 3 is an explanatory diagram of an operation expression of a vector space model.

FIG. 4 is an explanatory diagram of an operation of an information filtering unit according to the first embodiment of the present invention.

FIG. 5 is an explanatory diagram showing classification of users.

FIG. 6 is an explanatory diagram showing the order of filtering in Embodiment 1 of the present invention.

FIG. 7 is a configuration diagram of an information filtering device according to Embodiment 2 of the present invention.

FIG. 8 is a flowchart of a main part of an information filtering method according to Embodiment 2 of the present invention;

FIG. 9 is an explanatory diagram of a selection process of an information filtering unit in a specific example 2 of the present invention.

FIG. 10 is an explanatory diagram showing the order of filtering in Embodiment 2 of the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 User information storage part 2 Personal information management part 3 Interface management part 4 Information filtering part 7 User evaluation information storage part 8 Social filtering part 10 Information source 20 User

Claims (5)

[Claims] 1. An information resource adapted to a profile of a user is collected, another user using the same profile in a different meaning is extracted, and a correlation between the other user and the user is obtained. If there is no information resource, the information resource of interest to the other user is excluded from the collected information resource to obtain a result of filtering for the user. 2. Collecting information resources conforming to the user's profile and outputting them to the user. As a result, among the output information resources, the user is interested in information resources that the user is not interested in, And extracting other users using the same profile as the user, and then determining the correlation between the user and other users,
When there is no correlation, an information filtering method in which the information resource of interest to the other user is excluded from the collected information resource and is set as a new filtering result for the user. 3. The information filtering method according to claim 1, wherein the correlation between the user and another user is determined by performing social filtering. 4. A user information storage unit that stores attribute information of a plurality of users, a personal information management unit that manages information indicating a correlation with a plurality of user attributes stored in the user information storage unit, The information resource including the profile input by the user is output as a filtering result, and if there is an information resource that responds that the specific user is not interested in the filtering result, the user is interested in the information resource. And extracting another user using the same profile as the specific user, inquiring of the personal information management unit whether the other user has a correlation with the specific user, If there is a response indicating that the other user is less correlated, the information resource including the profile of the specific user indicates that the other user is interested. Information filtering apparatus according to claim exclude information resources and some that an information filtering unit for a new filtering results to the specific user. 5. A user evaluation information storage unit for storing information evaluated by an arbitrary user for a specific information resource, a user who is a requester of information filtering, and a plurality of other users who share the same information resource. And a social filtering unit for extracting a user having the lowest similarity to the requester user among the plurality of other users based on the evaluation value, and a profile input by a specific user. Is output as a filtering result, and if there is an information resource that responds that the specific user is not interested in the filtering result, the user is interested in the information resource, and The other users using the same profile as the user are extracted, and among the other users, the user having the lowest similarity with the specific user is extracted. The social filtering unit is inquired of who the user is, and the information resource that the lowest similarity user responded to by the social filtering unit is interested in is excluded from the information resource including the profile of the specific user. An information filtering device, comprising: an information filtering unit that generates a new filtering result.