JP4745993B2 - Consciousness system construction device and consciousness system construction program - Google Patents

Description

The present invention relates to Consciousness System construction equipment you and Consciousness System building program, in particular, the sales system of content for service providers to sell (shelf allocation), by feeding back the user preference mining results, while along the user preference The present invention is suitable for application to a method for constructing a user-friendly content system.
In addition, shelf allocation is a way to classify and organize a wide variety of products according to themes such as purpose, function, design, price range, etc., so that shoppers can easily see, pick and select them, and effectively related products. Saying to determine the display position in combination.

2. Description of the Related Art In recent years, a service that uses a user context such as a user's interest and provides appropriate content to the user via a network has been attracting attention.
In addition, blog services that promote user discussions on the Internet and promote discussions between users, and social network services that discuss interests among users who have approved each other as friends are attracting attention. It is considered that the number of users who use such services and the services that use such services will increase in the future.

In addition, Amazon and Last. A service that automatically constructs a user's interest as a user profile based on the purchase and playback history of the product such as fm, and recommends the product based on the user profile has appeared, and various interest information sent by the user There is a growing interest in services that provide extremely detailed information.
And this kind of information distribution service expands the interests of each user by browsing the articles and music information sent by users who are close to interest and the discussion contents in the community. Have the potential to become a foundation.

In Non-Patent Document 1, a user profile that holds not only a keyword that the user is interested in but also information that is the background of the keyword is automatically constructed, and a blog search is performed based on the user profile. A method for providing functionality is disclosed.
That is, in Non-Patent Document 1, by classifying articles accumulated by users into service ontology provided by service providers such as music and movies, the instance subject to the topic and background knowledge of the instances are automatically applied to the article. An auto-tagging method for automatically tagging and an interest ontology generation method for automatically extracting, as a user profile, an interest ontology in which a user's interest is class-hierarchized are disclosed.

Then, the degree of approximation between the interest ontology is measured, a class whose class hierarchy is different between the interest ontology with a high degree of approximation is detected, and an article belonging to that class is recommended as an unexpected interesting article to the user. A verification experiment is being conducted on an actual blog site to expand the range of interest and promote communication with other users.
This method is effective for content search and recommendation using exhaustive and specialized knowledge based on a specialized content classification system (referred to as taxonomy) created by a service provider.

  On the other hand, a folk sonomi that builds a relationship between contents based on tags given by users to contents on a network is expected as a method for constructing a user-driven service ontology. Folk sonomis have the potential to organize a vast amount of content based on the user's intuition, and are expected to help find content that matches the user's consciousness.

In addition, folk sonomi is a system for searching by content contributors such as photos, videos, documents, etc. by freely adding or classifying tags to content posted on Internet websites. Say to help. In this folksonomy, unlike the conventional classification / search method, in which the author or administrator of the site himself / herself classifies the content without involving the viewer, the classification work is open to all people and is an open one that never ends. It is characterized by collaboration. A tag is an index word, keyword, or label that describes the content and attributes of content such as “cat pictures”, “fishery”, and “artificial intelligence” in a single word. Tagging is called tagging.
Makoto Nakatsuji, Yu Miyoshi, Yoshihiro Otsuka: “A Blog Community Formation Method by Extracting User Interest Ontologies”, Database Society of Japan Letters, Vol. 5, no. 1 (2006)

  However, the information search in the current blog service is only a keyword-based search using a description such as a Web page search engine such as Google or RDF Site Summary (RSS). There was a problem that it was difficult to properly select an article that posted information according to one's own interest from the article.

  That is, each time a user performs a blog search, the user needs to appropriately configure a search object that matches his or her interests, and it takes time to select a search keyword. In addition, if the search target is not understood to some extent, the keyword itself cannot be configured. If the keyword cannot be specified even though it may be interested, the information search itself cannot be performed. It was.

  Further, in the method disclosed in Non-Patent Document 1, a specialized content classification system created by a service provider is adopted, which is easy for the service provider to manage, but the classification system is different from the system imaged by the user himself / herself. Therefore, there was a problem that intuitive access to the content was not possible. In addition, in such a content classification system, since an increase in access from users cannot be expected, there is a problem that it does not contribute much to an increase in sales of content.

In addition, the relationship between tags obtained with conventional folk sonomis is effective for constructing equivalence and contradiction relationships, but a method for constructing ontologies for expressing specialized knowledge such as inheritance relationships between classes. There was a problem that was not suitable. In addition, there is a problem that only content tagged by the user is targeted, and content that the user does not grasp is not considered.
It is an object of the present invention, while maintaining the content system that utilizes a comprehensive and specialized knowledge, your consciousness system construction equipment that is capable of building an intuitive content system of consciousness of the user is reflected And provide a system for building awareness systems.

In order to solve the above-described problem, according to the consciousness system construction device according to claim 1, an approximation of interest between users is obtained by measuring an approximation between interest ontologies in which personal interest information is conceptually hierarchized. By dividing the user network by the degree of approximation measuring means for measuring the degree, the user network generating means for generating a user network by establishing a link between the users whose degree of approximation of interest is within a predetermined range, a consciousness system construction apparatus and a network dividing means for layering awareness of the user, the network dividing means, as a network division method, dark density of population internal links to the link of the current Dhamma -out based on the score indicating an index modularity increases representing a is, by using the Newman method so as to determine the split edges It is adapted to perform a division of the network, for the combination of the first splitting edge randomly determined regardless the score Upon and performs the division of the network, to the network based on a combination of the determined edge performs first division for, in due Ri resulting cluster set to the division of the first said, said in the following cases a certain threshold number of users belonging to each cluster, different combinations a combination of split edges I line the first division again to change to, between the number of users belonging to each cluster is less than or equal to the threshold, repeated the first division, obtained Ri I said in the first division in was cluster set further second division for the cluster number of users exceeds the threshold value for the membership Performed, and the number of users belonging to the cluster and performs repeatedly the second division relative to the cluster until equal to or smaller than the threshold value.
According to the consciousness system construction device according to claim 2, the interest ontology is generated by extracting classes or instances included in the template ontology in which the interest information is conceptually hierarchized and classes higher than those classes. It is characterized by.

Further, according to the consciousness system construction device according to claim 3, the template ontology is configured such that words are conceptually hierarchized as the interest information, and the interest ontology is obtained from the template ontology in which the words are conceptually hierarchized. It is formed by extracting a hierarchical structure that contains the class or instance and class of their upper contain words as interesting information of the individual, the degree of approximation measuring means, class containing words included in the model ontology is the based on whether the determination result included in the interest ontology, with measures the degree of approximation topology between the interest ontology, determination instance containing words included in the model ontology of whether contained in the interested ontology Based on the results, measure the closeness of the classes between the ontology of interest, Based on the similarity of the serial class, characterized by measuring the degree of approximation between the interest ontology.

According to the consciousness system construction program according to claim 4, the morphological analysis is applied to the blog entry to extract the words included in the blog entry, and the words are conceptually hierarchized as the interest information. Selecting a template ontology, extracting a class or instance including a word extracted from the blog entry from the template ontology, and using the extracted class or instance and its higher class as an interest ontology. A step of extracting from a template ontology, a step of measuring the degree of interest between users by measuring the degree of approximation between the interest ontology, and a link between users whose degree of approximation of interest is within a predetermined range To create a user network , As the network division method, based-out to score an indication that modularity is increased which represents the density of the density of population internal links to the link of the current Dhamma, Newman method so as to determine the edges to divide the The network is divided using the network, and the first combination of edges to be divided in dividing the network is determined randomly regardless of the score, and based on the determined combination of edges. performs first division to the network, the I Ri resulting cluster set to the division of the first said, if less than the threshold value with the number of users belonging to each cluster, the combination of the edge to split the performing the first divided again changed to a different combination, belonging to each cluster yOU During the number of it is less than or equal to the threshold, repeated the first divided, said at first by Ri obtained cluster set the division, for the cluster number of users the membership exceeds the threshold further It performs second division, and by the number of users belonging to the cluster repeated the second division relative to the cluster until less than or equal to the threshold, the computer and the step of layering the awareness of the user It is made to perform.

  As described above, according to the present invention, it is possible to generate an interest ontology that reflects an individual's interest while inheriting the topology possessed by the template ontology, and the degree of approximation between the interest ontologies is reflected. By dividing the created user network, it is possible to map the template ontology created by the service provider and the user's consciousness, while maintaining the content system using comprehensive and specialized knowledge, while maintaining the user's consciousness. A reflected intuitive content system can be constructed.

Hereinafter, a consciousness system construction device and method according to an embodiment of the present invention will be described with reference to the drawings.
FIG. 1 is a block diagram showing a schematic configuration of a system to which a consciousness system construction device according to an embodiment of the present invention is applied.
In FIG. 1, the service provider 1 holds a purchase history 1b of user contents and products, and also holds a user ID 1a corresponding to the purchase history 1b. In addition, the blog company 2 holds a blog 2b centered on a dated article page and a user ID 2a corresponding to the blog 2b. The blog 2b can be held for each blog entry. Here, the blog entry represents the minimum unit of articles in the blog, and can be provided for each day.

Further, the service provider 3 classifies the articles accumulated by the user into the service ontology, thereby holding a content system 3a in which the instance that is the subject of the article and the background knowledge of the instance are tagged. Yes.
In addition, the user preference extraction unit 4 is provided with a preference knowledge base 4a in which information related to individual preferences obtained through purchase history and blog mining is stored.

  The preference analysis engine 5 also includes an approximation degree measuring means 5a for measuring an approximation degree of interest between users by measuring an approximation degree between interest ontology in which personal interest information is conceptually hierarchized. By creating a link between users whose degrees are within a predetermined range, the network generation means 5b for generating the user network, and the user network generated by the network generation means 5b are divided to stratify the user's awareness. A network dividing means 5c is provided.

The service providers 1 and 3, the blog operator 2, the user preference extraction unit 4, and the preference analysis engine 5 can be connected via a network such as the Internet.
Then, the preference analysis engine 5 constructs the user's interest ontology based on the service ontology, constructs the user's consciousness system by dividing the user network generated based on the degree of approximation between the interest ontology, As a result, a mapping between user awareness and service ontology can be obtained.

FIG. 2 is a diagram illustrating a relationship between user awareness and service ontology according to an embodiment of the present invention.
In FIG. 2, if the service provider 3 provides a service related to music to the user, the music service database 11a can hold a service ontology 12 related to music as the content system 3a.
Then, the preference analysis engine 5 constructs an interest ontology of the user 14 by using the service ontology 12a as a template ontology. Then, a user network is generated based on the degree of approximation between the interest ontologies of the user 14, the user network is divided and the consciousness of the user 14 is hierarchized, and the consciousness system 13 of the user 14 is constructed. As described above, the mapping between the consciousness 13a to 13c of the user 14 and the classes 12a to 12c constituting the service ontology can be obtained.

  The template ontology is a template of an interest ontology in which personal interest information is conceptually hierarchized. The template ontology can be arbitrarily created on the service provider 3 side. For example, when the service provider 3 holds a content system 3a related to music, a template ontology related to music can be constructed. Here, in order to express the user's interest in detail, it is preferable to construct a template ontology that is subdivided as much as possible and has high coverage. The entity of the template ontology is a representation of a class / instance tree on the DB, such as a text file described in an XML language such as the ontology description language OWL. In order to simplify the organization of information, instances may be classified only into the lowest class.

  This makes it possible to generate an interest ontology that reflects individual interests while inheriting the topology held by the template ontology, and to divide the user network that reflects the degree of approximation between interest ontologies. The service ontology 12a created by the service provider 3 and the consciousness 13a to 13c of the user 14 can be mapped, and the consciousness of the user 14 is reflected while maintaining the content system 3a using comprehensive and specialized knowledge. It is possible to construct an intuitive content system 3a.

Hereinafter, the operation of the preference analysis engine 5 in FIG. 1 will be described in more detail.
When the preference knowledge base 4a extracted by the user preference extraction unit 4 and the content system 3a possessed by the service provider 3 are input, the preference analysis engine 5 in FIG. 1 is based on the service ontology constituting the content system 3a. By performing auto-tagging on the article, it is possible to generate a user's interest ontology.
Here, the preference analysis engine 5 may generate an interest ontology in which individual interest information is conceptually hierarchized by extracting classes or instances included in the service ontology as a template ontology and higher classes thereof. it can.

  In other words, the preference analysis engine 5 extracts morphemes that frequently appear in a plurality of blog entries of the same user by applying morphological analysis to each blog entry input as the preference knowledge base 4a. Then, each morpheme that frequently appears in the blog entry is applied to the template ontology to check whether there is a character string that matches the class or instance in the template ontology. If there is a character string that matches the class or instance in the template ontology, the direct descendant classes and instances up to that class or instance can be extracted as the interest ontology from the root class of the template ontology.

Then, when the interest ontology of the user 14 is generated, the approximation degree measuring unit 5a measures the degree of interest between the users 14 by measuring the degree of approximation between the interest ontology.
Here, the degree-of-approximation measuring means 5a can measure the degree of approximation of the topology between the interest ontologies based on the co-occurrence degree between the class sets forming the topology of the interest ontology of the user 14. For example, the degree-of-approximation measuring means 5a measures the degree of approximation of the topology between the interest ontology based on the value obtained by dividing the number of members of the intersection set between the classes of interest ontology by the number of members of the class set of the template ontology 6. can do.

  Further, the degree-of-approximation measuring means 5a can measure the degree of approximation of classes between interest ontologies based on the degree of co-occurrence between instance sets included in the user 14's interest ontology. For example, the degree-of-approximation measuring means 5a measures the degree of approximation of the class between the interest ontology based on the value obtained by dividing the number of members of the product set between the instance sets of interest ontology by the number of members of the instance set of template ontology. be able to. Then, the degree-of-approximation measuring means 5a can measure the degree of approximation between the interest ontology by comprehensively determining the degree of approximation of the topology between the interest ontology and the degree of approximation of the class between the interest ontology.

FIG. 3 is a diagram showing a method of measuring the degree of approximation between interest ontology according to an embodiment of the present invention. In the following description, when the degree of approximation between an ontology and another ontology is measured, the former is called a source ontology and the latter is called a target ontology.
In FIG. 3, classes “b1”, “b2”, and “b3” exist immediately below the class “a1” of the template ontology OH, and “c1” and “c1” immediately below the class “b1”. It is assumed that a class called “c2” exists and classes “d1” and “d2” exist immediately below the class called “c1”. Also, it is assumed that classes “c3” and “c4” exist immediately below the class “b2”, and a class “c5” exists immediately below the class “b3”.

  The class “d1” includes instances “j” and “k”, the class “d2” includes the instance “l”, and the class “c2” includes “m”. "B", the class "b2" has an instance "n", and the class "c3" has "a", "e", "c", "f", "b" ”,“ D ”, and“ g ”exist, and the class“ c4 ”includes instances“ p ”,“ g ”,“ j ”, and“ h ”.

Then, by extracting words frequently appearing in each user's blog entry, and extracting the classes or instances containing the words and all the classes above them from the template ontology OH, interest ontologies OA and OB are created. Shall be.
Here, there are “b1” and “b2” classes immediately below the “a1” class of the interest ontology OA, and “c1” and “c2” classes immediately below the “b1” class. It is assumed that a class “d1” exists immediately below the class “c1”, and a class “c3” and “c4” exist immediately below the class “b2”. , The class “d1” includes instances “j” and “k”, the class “c2” includes the instance “m”, and the class “c3” includes “a”. , “C”, “b”, and “d” exist, and the class “c4” includes instances “q” and “h”.

  In addition, classes “b1”, “b2”, and “b3” exist immediately below the class “a1” of the interest ontology OB, and a class “c1” exists immediately below the class “b1”. Assume that a class “d1” exists immediately below the class “c1”, and a class “c3” and “c4” exist immediately below the class “b2”. The class “d2” has an instance “l”, the class “b2” has an instance “n”, and the class “c3” has “a” and “c”. ”,“ E ”, and“ f ”exist, and“ c4 ”class includes“ p ”and“ j ”.

  Then, the common class excluding the terminal class is analyzed between the template ontology OH and the interest ontology OA and OB, and the topology composed of the parent and child classes with the common class as the parent class is extracted. In the example of FIG. 3, the end classes are “d1”, “d2”, “c3”, and “c4”. As a result, a child class set G1 having the class “a1” as the parent class, a child class set G2 having the class “b1” as the parent class, and the class “c1” being the parent class between the template ontology OH and the interest ontologies OA and OB. Child class set G3, child class set G4 having class “b2” as a parent class, and child class set G5 having class “b3” as a parent class can be extracted.

The analysis of the common class only needs to confirm whether the same class ID exists between the template ontology OH and the interest ontologies OA and OB. For this reason, it is not necessary to measure the degree of approximation of class name attributes, instance set properties, etc., and the amount of calculation can be reduced while maintaining the correspondence between classes accurately.
Next, the degree of approximation between the child class sets X and Y forming each topology between the interest ontology OA and OB is calculated with depth priority. Here, if the set of child classes of the classes constituting the template ontology OH is Z, the degree of approximation between the child class sets X and Y can be obtained by | X∩Y | / | Z |. By adding the approximation degree between the child class set for each topology, it is possible to measure the degree of approximation S _T interested ontology OA, between OB topology.

  For example, in the child class set G1, since the template ontology OH includes child classes “b1”, “b2”, and “b3”, the number of members of the child class set of the template ontology OH in the child class set G1 is 3 In the child class set G1, the interest ontology OA includes child classes “b1” and “b2”, and the interest ontology OB includes child classes “b1”, “b2”, and “b3”. Therefore, the child classes that are commonly included in the interest ontologies OA and OB are only “b1” and b2 ”, and the number of members of the product set of the child class sets of the interest ontologies OA and OB in the child class set G1 is 2. As a result, the degree of approximation between the child class sets G1 of the interest ontologies OA and OB is 2/3.

Similarly, the degree of approximation between child class sets G2 of interest ontology OA and OB is 1/2, the degree of approximation between child class sets G3 is 0/2, and the degree of approximation between child class sets G4 is 2/2. As a result, interest ontology OA, similarity _{S T} topology between OB becomes 2/3 + 1/2 + 0/2 + 2/2.
In the calculation of the approximate degree of topology between the interest ontologies OA and OB, the class ID assigned from the template ontology OH is referred to, and the one along the connection form of the template ontology OH exists between the interest ontologies OA and OB. You only have to confirm whether you want to do it. For example, whether or not the connection form a1-b1-c1 in the interest ontology OA, OB is the same as the connection form of the template ontology OH depends on whether the classes “a1”, “b1”, and “c1” are in the interest ontology OA, OB. It can be determined simply by confirming whether or not it is held at.

For this reason, in order to check the degree of coincidence of the topology between the interest ontologies OA and OB, it is only necessary to check the class ID of the interest ontologies OA and OB. Therefore, it is possible to reduce the calculation amount of the degree of approximation of the topology between the interest ontologies OA and OB.
Next, the degree of approximation between the common classes of interest ontology OA and OB is calculated. Here, when calculating the degree of approximation between the common classes of the interest ontologies OA and OB, an instance set belonging to the class can be used. That is, in a certain class C1, assuming that the instance set of the source ontology is x, the instance set of the target ontology is y, and the instance set of the template ontology OH is z, the degree of approximation between the common classes of the interest ontology OA and OB is | x ∩y | / | z |. By adding the degree of approximation between the common class, it is possible to measure the degree of approximation S _C between interest ontology OA, the OB classes.

  For example, in the common class “b2” between the interest ontologies OA and OB, the instance set G6 in the common class “b2” of the interest ontology OA includes the instance “n”, and the common class “b2” of the interest ontology OB An instance “n” exists in the instance set G6, and an instance “n” exists in the common class “b2” of the template ontology OH. As a result, the number of members of the common class “b2” of the template ontology OH is 1, the number of members of the product set of the instance set G6 of the interest ontologies OA and OB is 1, and the degree of approximation between the common classes “b2” is 1/1.

  Further, in the common class “c2” between the interest ontology OA and OB, the instance set G7 in the common class “c2” of the interest ontology OA includes instances “a”, “c”, “b”, and “d”. In the instance set G7 in the common class “c3” of the interest ontology OB, instances “a”, “c”, “e”, and “f” exist, and in the common class “c3” of the template ontology OH “A”, “e”, “c”, “f”, “b”, “d”, and “g” exist.

As a result, the number of members of the common class “c3” of the template ontology OH is 7, the number of members of the product set of the instance set G7 of the interest ontologies OA and OB is 2, and the degree of approximation between the common classes “c3” is 2/7.
Therefore, the degree of approximation _{S C} between interest ontology OA, the OB class is 1/1 + 2/7.
In the calculation of the degree of approximation between the common classes of the interest ontology OA and OB, it is only necessary to confirm whether or not the instance ID allocated from the template ontology OH exists between the interest ontology OA and OB. For this reason, in order to calculate the degree of approximation between the classes of interest ontology OA and OB, it is not necessary to confirm in advance the correspondence between instances based on the consistency of the names of instances, and the amount of calculation can be reduced.

The interest ontology OA, the similarity S _C between the degree of approximation S _T and class topology in OB is determined, by using an evaluation function f (X) corresponding to the importance for the topology and class, the following (1 ), The degree of approximation S _O (AB) between the ontology OA and OB of interest can be given.
S _O (AB) = S _T + f (S _C ) (1)
This makes it possible to construct the interest ontology OA and OB while inheriting the class characteristics of the template ontology OH, and directly measure the degree of co-occurrence between instance sets in the degree of approximation between the interest ontology OA and OB. It is possible to appropriately extract interest ontologies OA and OB adapted to individual preferences while suppressing the amount of calculation, and at the same time, the knowledge ontologies OA and OB have knowledge of the domain of the template ontology OH. It can be used as a relative measure of possession, and users who have a lot of knowledge about the domain can be effectively narrowed down.

Also, when measuring the degree of approximation S _T topology interest ontology OA, the OB, based on the number of members of the class set stationery ontology interest ontology OA, by measuring the degree of approximation topology between OB, interest ontology Even when the number of members of the class set of OA and OB increases, it is possible to prevent the degree of approximation of the topology between the interest ontology OA and OB from being reduced, and there is a large amount of knowledge in the sense that there are abundant classes. The degree of approximation with the interest ontology OA, OB can be increased.

FIG. 4 is a diagram illustrating the superiority of the method for measuring the degree of approximation between interest ontology according to an embodiment of the present invention.
In FIG. 4, in ontology OC, it is assumed that there is no class “b3” in ontology OB in FIG. In this case, in ontology OB, child classes “b1”, “b2”, and “b3” are included in child class set G1, and in ontology OC, child classes “b1” and “b2” are included in child class set G1. Although it is included, the degree of approximation between the child class sets G1 of the interest ontologies OA and OC is 2/3, similar to the degree of approximation between the child class sets G1 of the interest ontologies OA and OB.

  As a result, in the ontology OC, although there is no class “b3” of the ontology OB in FIG. 3, the degree of approximation between the interested ontology OA and OC is larger than the degree of approximation between the interested ontology OA and OB. The degree of approximation with an interest ontology having a large amount of knowledge can be prevented from becoming relatively small in the sense that there are abundant classes.

Next, in FIG. 1, when the degree of approximation between the interest ontologies of the users 14 is measured by the degree-of-approximation measuring means 5 a, the network generation means 5 b reads between the users 14 whose interest degrees of interest are within a predetermined range. Create a user network by creating a link.
That is, the network generation unit 5b considers that the instance set shared by the user sets having similar approximations between the interest ontologies reflects the consciousness of the user set, and extracts such instance sets. Therefore, by applying equation (1), the degree of approximation S (ij) of the user j viewed from the user i is obtained, and when the condition X> S (ij)> Y is satisfied, the interest between the users i and j As a result, a relationship is established between users i and j. Then, as shown in FIG. 5, a user network is generated by associating the users i and j with the nodes and establishing a link between the users i and j with close interest.

Here, a user with a very large S (ij) is a user who has a wide range of interest. For this reason, when a link is established between users i and j with close interest, by setting the condition X> S (ij), it is possible to remove users who have a wide range of interests and inhibit network division. Can be prevented.
Note that, when a relationship is established between the users i and j, a link may be established only between users of the top Z (Z is a positive integer) viewed from the user i.
Next, in FIG. 1, when the user network is generated by the network generation unit 5b, the network dividing unit 5c divides the user network generated by the network generation unit 5b, thereby layering the user's consciousness. To do.

FIG. 6 is a diagram illustrating a method for dividing a user network according to an embodiment of the present invention.
In FIG. 6, assuming that the user networks for the users A to J are generated by the network generation unit 5 b of FIG. 1, the network dividing unit 5 c divides the user network so that the consciousness of the users A to J is obtained. Generates user sets Q1 to Q3 which are hierarchized.
Here, as a method of dividing the user network, as described in “Newman, MEJ: Fast algorithm for detecting community structure in networks (2003)”, a link inside a group with respect to a link between groups. It can be evaluated based on the density of the density. This degree is defined by a value called modularity Q.

Here, if the variation in the number of users belonging to the user set generated after being divided by the method described in the above literature is large or the number of users belonging to each user set is too small, The user set hierarchy cannot be obtained. That is, in the method described in the above document, when determining the edge to be divided first, if there are a plurality of combinations of edges having the same value of the score ΔQ indicating the index that increases the modularity Q after the combination, the combination of edges Is decided at random. For this reason, the variation of the number of users belonging to the user set generated after being divided by the method described in the above literature becomes large, and the genre system is divided into a part that is too deep and a part that is too shallow. This makes it difficult to handle the hierarchy of user sets.

Therefore, the network dividing unit 5c randomly determines a combination of edges to be divided first. Then, in the cluster (set) that is finally divided first, the division is permitted only when the number of users belonging to the second and later divided clusters exceeds a certain threshold θ, and the second and later When the number of users belonging to the cluster to be divided is equal to or smaller than the threshold θ, the combination of edges to be divided first can be changed.

  The first divided cluster often has a very large number of users belonging to the cluster, and is considered not to be a fine user set. Therefore, the division can be repeated until the cluster is below the threshold θ. And the cluster divided | segmented in this way can be made into the artist set along one consciousness. Thereby, variation in the number of users belonging to the divided clusters can be suppressed, and the granularity of the hierarchical awareness can be equalized.

If a cluster exceeding the threshold θ is not generated even if the division is repeated a plurality of times, the number of links in the user network is reduced by reducing the parameters X and Z, and the user network is arranged so that the division is possible. Can do.
Then, the network generation unit 5b can re-divide the divided cluster in which the number of users exceeds the threshold θ by changing the threshold θ, and can hierarchize the consciousness 13a to 13c of the user 14.
The preference analysis engine 5 can feed back the hierarchical consciousness 13a to 13c of the user 14 to the content system 3a held by the service provider 3 as a user preference mining result.

  As a result, it is possible to reconstruct the content system 3a that is easy to understand for the user 14 according to the preference of the user 14, and obtain a mapping between the comprehensive and specialized content system 3a desired by the service provider 3. be able to. For this reason, the user 14 can discover content intuitively, and can efficiently discover a product in line with his / her interest. In addition, the service provider 3 can manage the product while maintaining the content system 3a that is easy to manage, making it easier to market the trend of interest of the user 14, and accessing the content system 3a from the user 14 It is possible to promote the sales of the product.

FIG. 7 is a diagram illustrating a community formation method using the approximation of interest ontology according to an embodiment of the present invention.
In FIG. 7, it is assumed that the blog entries PA and PB of the users A and B are classified with respect to the template ontology, thereby generating the interest ontologies KA and KB of the users A and B, respectively (step S1). . Then, the degree of approximation between the interest ontologies KA and KB of the users A and B is measured (step S2), and the topologies are analyzed by analyzing classes and instances that co-occur between the interest ontologies KA and KB having a high degree of approximation. Although it is different, information that is highly likely to be interesting can be recommended to the user as unexpected information through the other user's entry (step S3).

For example, by measuring the degree of approximation between the interest ontologies KA and KB of each user A and B, a user who is interested in a class such as “Madchester” or an instance such as “Happy Mondays” can have a class “Glasgow” It can be seen that there is a high possibility of being interested in an instance of “Teenage Fanclub”.
Moreover, by applying such interest ontologies KA and KB to a blog, it is possible to support the formation of a community based on an unexpected entry recommendation based on the degree of approximation between the interest ontologies KA and KB rather than a simple keyword search. The user's interest can be naturally expanded (step S4).

The approximation degree measuring unit 5a, the network generating unit 5b, and the network dividing unit 5c can be realized by causing a computer to execute a program in which an instruction for performing processing performed by these units is described.
Then, if this program is stored in a storage medium such as a CD-ROM, the degree-of-approximation measuring means 5a, network generation is performed by installing the storage medium in the computer of the preference analysis engine 5 and installing the program in the computer. The processing performed by the means 5b and the network dividing means 5c can be realized. Moreover, this program can be easily spread by downloading this program via a network.

  Further, when a computer executes a program in which an instruction for executing the processing performed by the approximation degree measuring unit 5a, the network generating unit 5b, and the network dividing unit 5c is executed, it may be executed by a stand-alone computer. A plurality of computers connected to the computer may be distributed.

  The present invention can construct an intuitive content system that reflects the user's consciousness while maintaining a content system that uses comprehensive and specialized knowledge, and impairs the efficiency of product management by a service provider. Therefore, it is possible to promote the user's access to the content system held by the service provider, and increase the sales of the product.

It is a block diagram which shows schematic structure of the system with which the consciousness system construction | assembly apparatus which concerns on one Embodiment of this invention is applied. It is a figure which shows the relationship between a user's consciousness and service ontology concerning one Embodiment of this invention. It is a figure which shows the approximation degree measuring method between the interest ontology concerning one Embodiment of this invention. It is a figure which shows the predominance of the approximation degree measuring method between interest ontology concerning one Embodiment of this invention. It is a figure which shows the method of networking the interest relationship between the users based on one Embodiment of this invention. FIG. 3 is a diagram illustrating a method for dividing a user network according to an exemplary embodiment of the present invention. It is a figure which shows the community formation method using the approximation of the interest ontology which concerns on one Embodiment of this invention.

Explanation of symbols

1, 3 Service provider 1a, 2a User ID
1b purchase history 2 blog operator 2b blog 3a content system 4 user preference extraction means 4a preference knowledge base 5 preference analysis engine 5a approximation measure means 5b network generation means 5c network division means

Claims (4)

Approximation degree measuring means for measuring an approximation degree of interest between users by measuring an approximation degree between interest ontology in which personal interest information is conceptually hierarchized;
User network generation means for generating a user network by establishing a link between users whose interests are in a predetermined range; and
A network division unit that divides the user's consciousness by dividing the user network, and a consciousness system construction device comprising:
The network splitting means and the network division method, based-out to score an indication that modularity is increased which represents the density of the density of population internal links to the link of the current Dhamma, to determine the dividing edges The newman method is used to divide the network, and the combination of edges to be divided first in dividing the network is determined randomly regardless of the score. performs first division to the network based on a combination of edges, in the resulting cluster aggregate that in the first division, if: there is a number of user thresholds belonging to each cluster, wherein the combination of the divided edges again changed to a different combination I 1 th row split, the During the number of users is less than or equal to the threshold value belonging to the cluster, it repeated the first divided,
Wherein the first by Ri obtained cluster set the division, the number of affiliated user performs a further second division for the clusters that exceed the threshold, number the threshold value of the user and belonging to the cluster A consciousness system construction device, characterized in that the second division is repeatedly performed on the cluster until:   The consciousness system construction device according to claim 1, wherein the interest ontology is generated by extracting classes or instances included in the template ontology in which the interest information is conceptually hierarchized and higher classes thereof. The template ontology is formed by conceptually hierarchizing words as the interest information, and the interest ontology includes classes or instances including words as the personal interest information from the template ontology in which the words are hierarchized conceptually, and those It is formed by extracting a hierarchical structure that includes classes above
The degree-of-approximation measuring means measures the degree of topology approximation between the interest ontologies based on a determination result of whether or not a class including a word included in the template ontology is included in the interest ontology, and Based on the determination result of whether or not an instance including the word included in the interest ontology is included in the interest ontology, the degree of approximation of the class between the interest ontologies is measured, and based on the degree of approximation of the topology and the degree of approximation of the class The consciousness system construction device according to claim 2 , wherein the degree of approximation between the interest ontology is measured. Extracting words contained in the blog entry by applying morphological analysis to the blog entry;
Selecting a template ontology in which words are hierarchized as interest information;
Extracting from the template ontology a class or instance containing words extracted from the blog entry;
Extracting the extracted classes or instances and their superior classes as interest ontology from the template ontology;
Measuring the degree of interest between users by measuring the degree of approximation between the interest ontology; and
Creating a user network by creating a link between users whose approximation of interest is within a predetermined range; and
As the network division method,-out based on the score indicating an index modularity increases which represents the density of the density of population internal links to the link of the current Dhamma, use Newman method which is adapted to determine the split edges Then, the network is divided, and the combination of edges to be divided first in dividing the network is determined randomly regardless of the score, and based on the determined combination of edges, performs first division to the network, the I Ri resulting cluster set to the division of the first said, said in the following cases a certain threshold number of users belonging to each cluster, the combination of split edges different said combination again to change by performing a first division of users belonging to each cluster While BUSY is below the threshold, it repeated the first divided,
Wherein the first by Ri obtained cluster set the division, the number of affiliated user performs a further second division for the clusters that exceed the threshold, number the threshold value of the user and belonging to the cluster A consciousness system construction program characterized by causing a computer to execute a step of hierarchizing user consciousness by repeatedly performing the second division on the cluster until:

Images