JP4443999B2 - Information management apparatus and information management program - Google Patents

Description

  The present invention relates to an information management apparatus, an information management method, and an information management program, and is particularly suitable for application to a method of automatically mapping information sources expressed hierarchically.

With the spread of the Internet this year, information that can be obtained by individuals has increased dramatically, and in order to make it possible to easily access desired information, such information is managed hierarchically.
Here, when information is managed, a request for an information management method differs depending on the purpose of use of the information, the collected information, and the like. For this reason, a unique concept hierarchy is constructed by each administrator, and information is distributed and managed while allowing each administrator to maintain consistency of the concept hierarchy.

Also, for example, in Non-Patent Document 1, the concept hierarchy for managing various information is regarded as one information source, and the information possessed by each information source can be used mutually. Based on the above, a machine learning technique is disclosed in which similarity between each concept is identified and learned as a rule for adjusting a difference from other information sources.
Further, for example, Non-Patent Document 2 discloses an ontology time that is independently updated or customized by extending to an ontology description language that can describe range restrictions and negation in query approximation conversion for non-uniform ontology. A method for dealing with spatial and spatial spread is disclosed.

Further, for example, in Non-Patent Document 3, the similarity between instances obtained mechanically is determined in order to easily identify where information managed by others is located in the concept hierarchy possessed by the other person. Based on the above, a method is disclosed in which instances that are not shared between concept hierarchies are automatically shared, so that connections can be made between concept hierarchies that do not have the same instance.
Ryutaro Ichise, Hideaki Takeda, Shinichi Honda “Learning the rules of coordination between hierarchical knowledge”, Journal of Artificial Intelligence, Vol. 17, no. 3, pp. 230-238 (2002) Akamine, Satoshi Hiramatsu, Tetsuji Sato “Query Approximation Transformation for Non-uniform Ontologies” AIJ Research Materials (SIG-SW & ONT-A301-07) (2003) Masahiro Amagasaki, Hideaki Takeda, Ryutaro Ichise “Integration of Internet Directory Using Hierarchical Knowledge and Content Similarity” The 17th Annual Conference of the Japan Society for Artificial Intelligence, 1D4-07 (200)

  However, in the methods disclosed in Non-Patent Documents 1 and 3, mapping between concepts is performed based only on the similarity between concepts or between instances. For this reason, when humans analyze the mapping results obtained by these methods, even if the mapping results are close to each other in terms of vocabulary, many incorrect mapping results that cannot always be determined as meanings between concepts can be obtained. was there.

In addition, the method disclosed in Non-Patent Document 2 relates to the extension of ontology description language, and there is a problem that the accuracy of the mapping result between concepts cannot be improved.
An object of the present invention is to provide a while suppressing the burden on the human, the concept between mapping accuracy can be improved information management equipment Contact and information management program.

In order to solve the above-described problem, according to the information management apparatus of the present invention, the inter-concept mapping extraction unit that extracts the concepts that are approximate to each other included in the two information sources, and the inter-concept mapping mapping unit, are extracted. A topology mapping extraction means for extracting a partial information source that is approximated between partial information sources that are composed of concept sets having a parent-child relationship based on the connection form of the concepts that are similar to each other; A partial information source in which the topology mapping result extracted by the mapping extraction means is multiplexed is detected, and an error in the topology mapping result is detected based on the number of concepts included in the detected partial information source. Between the concepts extracted by the inter-concept mapping extraction means by removing the detected topology mapping result Characterized in that it comprises a topology mapping evaluating means for correcting the mappings results.

According to the information management apparatus of the present invention, the inter-concept mapping extraction unit includes an approximation degree measuring unit that measures the closeness between instances of the concepts included in the two information sources, respectively. .
Further, according to the information management apparatus of the present invention, the topology mapping extraction unit includes a mapping state determination unit that determines a mapping state of a concept that has a parent-child relationship with the mapped concept, and the inter-concept mapping extraction unit. Connection topology determination means for determining the approximateness of the extracted connection forms of the concepts that are similar to each other, and the topology mapping extraction means is based on the approximation of the connection form determined by the connection form determination means, It is characterized in that topology mapping is formed between partial information sources composed of concept sets having a parent-child relationship with each other.

Also, according to the information management program according to the present invention, the steps of performing mapping between concepts for the two sources of information, based on the parent-child relationship between the mapped concept, connection form between concepts that approximate one another Based on the approximation of the connection form between the concepts and the step of determining the closeness, a partial information source that is approximated between partial information sources composed of concept sets that are approximated by the connection form is extracted and topology Mapping, a step of detecting a partial information source in which the topology mapping result is multiplexed, a step of calculating the number of concepts included in the detected partial information source, and included in the partial information source based on the number of concepts, detects an error of the topology mapping result, the approximate by removing the detected topology mapping result of the error Characterized in that to perform the steps on a computer to modify the mapping result between.

  As described above, according to the present invention, mapping between concepts included in different information sources can be automatically performed in consideration of not only the similarity between concepts but also the hierarchical structure of the concepts. For this reason, it is possible to improve the mapping accuracy between concepts contained in different information sources while suppressing the burden on human resources, and where the information managed by others is located in the concept hierarchy Can be accurately identified. As a result, even when information is distributed and managed for each administrator, it is possible to efficiently access the information required by the administrator, and even for small-scale distributed and diverse system data, Flexible access, cooperation and management can be realized.

It will be described below with reference to the drawings attached to the information management equipment according to an embodiment of the present invention.
FIG. 1 is a block diagram showing a schematic configuration of an information management apparatus according to an embodiment of the present invention.
In FIG. 1, the information management apparatus includes an inter-concept mapping extraction apparatus 2, a topology mapping extraction apparatus 3, and a topology mapping evaluation apparatus 4. Here, the inter-concept mapping extraction device 2 extracts concepts that are included in the two information sources 1a and 1b and that are similar to each other. The topology mapping extraction device 3 performs topology mapping based on the concept connection forms extracted by the inter-concept mapping extraction device 2. Topology Mapping evaluation device 4, based on preparative Polo di mapping extractor topology mapping result extracted in 3, the mapping result evaluation between concepts extracted by the concept between mapping extractor 2.

  Here, the inter-concept mapping extraction apparatus 2 can be provided with an approximation degree measuring means 2a for measuring the closeness between the instances of the concepts included in the two information sources 1a and 1b. As a method for measuring the closeness between instances, lexical approximation in each instance, κ statistics, and the like can be used. When the degree of approximation between instances is equal to or more than a heuristic threshold, mapping between concepts included in the two information sources 1a and 1b can be performed. Note that the information sources 1a and 1b provide information every moment, and are configured by a hierarchical structure such as an ontology, an Internet directory, and a book catalog. Further, the information sources 1a and 1b may be content data such as images, and when measuring the closeness between images, the meta information added to the images may be used.

  Further, the topology mapping extraction device 3 is extracted by the mapping state determination means 3a for determining the mapping state of the concept having the parent-child relationship with the concept mapped by the inter-concept mapping extraction device 2, and extracted by the inter-concept mapping extraction device 2. In addition, it is possible to provide connection form determining means 3b for determining the closeness of the connection forms of concepts that are similar to each other. Then, the topology mapping extraction device 3 can form a topology mapping between partial information sources composed of concepts having a parent-child relationship based on the closeness of the connection form determined by the connection form determination unit 3b. it can. The partial information source refers to a part of the information sources 1a and 1b, and specific examples of the partial information source include a sub ontology and a sub directory.

The topology mapping evaluation apparatus 4 includes a partial information source detection unit 4a for detecting a partial information source that gives a multiplexed topology mapping result, and a concept included in the partial information source detected by the partial information source detection unit 4a. The concept number calculating means 4b for calculating the number of s and the topology mapping error detecting means 4c for detecting an error in the topology mapping result based on the number of concepts included in the partial information source calculated by the concept number calculating means 4b are provided. be able to.
The mapping is to determine whether the concept and the concept are semantically equal. The topology means a connection form of concept sets. The topology mapping is to determine whether concept connection forms are equal between concept sets.

FIG. 2 is a diagram illustrating an example of the structure of an information source according to an embodiment of the present invention.
In FIG. 2, for example, the hierarchical structure of the concept 11 included in the information source 1a can be represented by a tree structure, and an instance 12 is assigned to a node of the tree. Similarly, the information source 1b can be represented by the tree structure of the concept 11. The instance 12 is the actual state of information and is assigned to the concept 11.

FIG. 3 is a flowchart illustrating an information management method according to an embodiment of the present invention.
In FIG. 3, when the information sources 1a and 1b are input to the inter-concept mapping extraction apparatus 2 in FIG. 1 (step S1), the approximation degree measuring means 2a reads between the instances of the concepts included in the information sources 1a and 1b, respectively. Is measured (step S2). Then, the inter-concept mapping extraction device 2 performs mapping between the concepts included in the information sources 1a and 1b based on the closeness of the instances of the concepts included in the information sources 1a and 1b (step S3).

  When the mapping between concepts is performed by the inter-concept mapping extraction device 2, the mapping result is output to the topology mapping extraction device 3. When the mapping result between the concepts is output to the topology mapping extraction device 3, the topology mapping extraction device 3 is based on the closeness of the connection forms of the concepts that are approximated to each other extracted by the inter-concept mapping extraction means 2. Topology mapping is performed (step S4).

  Here, when topology mapping is performed, the mapping state determination unit 3a can determine the mapping state of the concept having a parent-child relationship with the concept mapped by the inter-concept mapping extraction apparatus 2. And the connection form determination means 3b can determine the closeness of the connection form between the concepts which approximate each other based on the parent-child relationship between the mapped concepts.

FIG. 4 is a diagram illustrating a topology mapping extraction method according to an embodiment of the present invention. In the following description, an ontology will be described as an example of data held by the information sources 1a and 1b.
4, the ontology O concepts in _A C _a1 -C _a3, the ontology O _B is assumed that there are concepts C _b1 -C _b3. Incidentally, ontology O _A, and O _B, information sources 1a, the data held by 1b, sources 1a, those defined as scheme based on the concept for 1b. The concepts C _{a1 to} C _a3 constitute a parent-child relationship, and the concepts C _{b1 to} C _b3 constitute a parent-child relationship.

Then, as shown in FIG. 4 (a), it is assumed that the concept C _b2 concepts C _a2 and ontology O _B ontology O _A in concept between mapping extraction device 2 of FIG. 1 are mapped to one another (M1 ). Then, when the concepts C _a2 and C _b2 mapped to each other exist in the ontology O _A and O _B , the topology mapping extraction device 3 determines whether or not a parent concept exists in these concepts C _a2 and C _b2. To do.

Then, as shown in FIG. 4B, when the parent concepts C _a1 and C _b1 exist in these concepts C _a2 and C _b2 , the topology mapping extraction device 3 performs the operation in the inter-concept mapping extraction device 2. By referring to the mapping result, it is confirmed whether or not the parent concepts C _a1 and C _{b1 of} these concepts C _a2 and C _b2 are mapped to each other. The concept C _a2, C parent concept C _a1 in _b2, if the C _b1 is mapped to each other (M2), the concept set {C _a1, C _a2} sub ontology O _SA1 and concepts set consisting of {C _b1, C _It is determined that the conceptual connection form is the same as that of the sub-ontology _OSB1 composed of _b2 }, and a topology mapping is formed between these sub-ontologies _OSA1 and _OSB1 (TM1).

Next, when the topology mapping is formed between the sub-ontologies _OSA1 and _OSB1 , the topology mapping extraction device 3 determines whether or not there are child concepts in the concepts C _a2 and C _b2 .
Then, as shown in FIG. 4C, when the concepts C _a2 and C _b2 have child concepts C _a3 and C _b3 , respectively, the topology mapping extraction device 3 performs the operation in the inter-concept mapping extraction device 2. By referring to the mapping result, it is confirmed whether or not the child concepts C _a3 and C _{b3 of} these concepts C _a2 and C _b2 are mapped to each other. The concept C _a2, C children concept C _a3 of _b2, if the C _b3 are mapped to one another (M3), the concept set _{{C a1, C a2, C} a3} sub ontology O _SA2 and concepts set consisting of {C _It is determined that the conceptual connection form is the same as that of the sub ontology _OSB2 composed of _b1 , _Cb2 , _Cb3 }, and a topology mapping is formed between the sub ontology _OSA2 and _OSB2 (TM2).

By repeating the above operations ontology O _A, between O _B, topology mapping extractor 3 extracts the topology mapping, inputs the topology mapping results in the topology mapping evaluation device 4. When the topology mapping result is input to the topology mapping evaluation device 4, the topology mapping evaluation device 4 extracts the inter-concept mapping extraction device 2 based on the topology mapping result extracted by the topology mapping extraction device 3. Evaluate the mapping results between the concepts.

  Here, when evaluating the mapping result between concepts extracted by the inter-concept mapping extraction apparatus 2, the partial information source detection unit 4a can detect a sub ontology that gives a multiplexed topology mapping result. . When the partial information source detection unit 4a detects the sub ontology, the concept number calculation unit 4b calculates the number of concepts included in the detected sub ontology (step S5). The topology mapping error detection unit 4c detects an error in the topology mapping result based on the number of concepts included in the sub ontology.

  Here, the topology mapping error detecting means 4c can detect, as an error, a topology mapping result formed from sub-ontologies other than the sub-ontology having the largest number of concepts constituting the sub-ontologies among the detected sub-ontologies. it can. Then, the topology mapping evaluation device 4 corrects the topology mapping result based on the number of concepts included in the sub ontology (step S6). And the correct mapping result 5 can be output by correcting the mapping result between concepts so as to follow the corrected topology mapping result.

For example, the concept between mapping, if the concept C _ai with ontology O _A in FIG. 4 has a plurality of concepts and mappings ontology O _B, partial information source detecting means 4a refers to the topology mapping results, In the ontology O _A of the ontology forming the topology mapping, the one containing the concept C _ai is detected. Then, among the detected sub-ontologies, topology mapping results other than the sub-ontology having the largest number of concepts constituting the sub-ontologies are detected as errors. Then, the inter-concept mapping result according to the finally obtained topology mapping result can be output as a correct answer.

  Thereby, the mapping between the concepts included in the different information sources 1a and 1b can be automatically performed while considering not only the similarity between the concepts but also the hierarchical structure of the concepts. For this reason, it becomes possible to improve the mapping accuracy between concepts included in different information sources 1a and 1b while suppressing the burden on manpower, and where the information managed by others is in the concept hierarchy of one's own It can be accurately identified whether it is located. As a result, even if the information 1a and 1b is distributed and managed for each manager, it becomes possible to efficiently access the information required by the manager, and the data possessed by a small and distributed system However, flexible access, cooperation, and management can be realized.

  For example, since automatic mapping based only on the closeness of the vocabulary between concepts was performed, even when the mapping result between concepts contains many errors, the hierarchical structure of the concept of the information source is not manually scrutinized, It becomes possible to correct the mapping result between the concepts correctly. For this reason, even when an administrator or user of a certain information source adds an instance of a concept determined to be equal in different information sources to an information source managed by himself or uses it for discovery of an instance, It is possible to prevent the addition of an instance that does not match the hierarchical structure of the information source of the user, and the user can prevent the discovery of the instance contrary to his intention.

In the embodiment described above, the information source 1a, ontology O _A as data having the 1b, although the O _B was described as an example, ontology O _A, in addition to O _B, Internet directory hierarchy such classification catalog of books The present invention may be applied to a method for automatically mapping information sources that are expressed in an automatic manner.
For example, two Internet directories can be mapped between concepts, instances belonging to the concepts determined to be the same concept can be moved to one of the Internet directories, and integrated and held as one Internet directory. Thereby, the search user can acquire the instance which two Internet directories have by one search, utilizing a concept hierarchy.

Note that the inter-concept mapping extraction device 2, the topology mapping extraction device 3, and the topology mapping evaluation device 4 in FIG. 1 can be realized by causing a computer to execute a program that performs processing performed in these devices.
Hereinafter, the Internet directories of “infoseek (registered trademark)” and “YAHOO! (Registered trademark)” will be referred to as information sources 21a and 21b, respectively, and “infoseek” and “YAHOO! A method of performing topology mapping between concepts will be specifically described by taking the case of “Internet directory category” as an example.

FIG. 5 is a diagram showing a configuration example of an Internet directory of “infoseek” and “YAHOO!”. In the example of FIG. 5, concepts below the UNIX (registered trademark) category of the Internet directory of “infoseek” and “YAHOO!” Are targeted.
In FIG. 5, the concept of “UNIX (registered trademark)” of “infoseek” includes child concepts of “Solaris (registered trademark)”, “Linux (registered trademark)”, and “FreeBSD (registered trademark)”. Yes. In addition, the concept of “Linux” includes child concepts of “Turboline (registered trademark)”, “Mkulinux (registered trademark)”, and “newspaper and magazine”.

  On the other hand, the concept of “UNIX (registered trademark)” of “YAHOO!” Has child concepts of “Linux”, “Solaris”, and “BSD (registered trademark)”. Further, the concept of “Linux” includes child concepts of “distribution” and “support and documentation”. In addition, the concept of “distribution” includes child concepts of “TurboLinux” and “Red Hat (registered trademark)”.

  Then, the inter-concept mapping extraction apparatus 2 in FIG. 1 performs mapping between the concepts included in each of the information sources 21a and 21b, with the category in FIG. 5 as a concept. Here, each category of the Internet directory has a set of pages. And the degree of approximation between the page sets which each category has can be used for mapping between the concepts included in the information sources 21a and 21b.

  For example, as shown in FIG. 6, when measuring the degree of approximation between the “UNIX (registered trademark)” concept of “infoseek” and the “Unix (registered trademark)” concept of “YAHOO!” The degree of approximation between the set of pages that the category has can be used. Here, the degree of approximation between page sets can be calculated using, for example, a vector space model. That is, all pages and categories can be expressed by weight vectors of the same search word, and the degree of approximation between documents can be calculated from the similarity between vectors.

The concept of “UNIX (registered trademark)” in “infoseek” and “YAHOO! If the degree of approximation S (UNIX (registered trademark), Unix (registered trademark)) between the concept of "Unix (registered trademark)" is greater than the heuristic threshold θ, "infox" (UNIX) (registered trademark) It can be determined that the concept “Unix” of “YAHOO!” Can be mapped.
That is, since a document can be characterized by several terms, considering a vector space in which individual terms correspond to each dimension, the document can be expressed as a point in the vector space. Here, the term frequency tf _ji , document frequency df _i, or tf _ji × idf _j can be used as the weight w _ji of the term t _i in the document D _j for obtaining the dimension in the dimension direction corresponding to each term. .

Term frequency tf _ji not represent either term t _i appears in how often the document D _j, can be defined by the following equation.
tf _ji = freq (i, j)
Here, freq (i, j) is the appearance frequency of the term t _i in the document D _j .
As a modification of the term frequency tf _ji , there is also a method of normalizing with a weight K and a maximum frequency max _{i, j as} shown in the following equation.
tf _ji = K + (K−1) freq (i, j) / (max _{i, j} freq (i, j))

Further, as a modification of the term frequency tf _ji , there is a method of normalizing with the log and the number of terms in the document as shown in the following equation.
tf _ji = log ₂ (freq (i, j) +1) / log ₂ (number of types of terms in document j)
The document frequency df _i indicates that the term t _i represents the number of documents, and can be defined by the following equation.
df _i = Dfreq (i)
However, Dfreq (i) is the number of documents in which the term t _i appears. In practice, this reciprocal idf _i is used.

As a modification of the idf _i, is as shown by the following equation, a method of normalizing the document total number N of log and document set in.
idf _i = log ₂ (N / Dfreq (i)) + 1
Further, tf _ji × idf _j has both the properties of the term frequency tf _ji and the document frequency df _i , and the weight w _ji of the term t _i in the document D _j can be defined by the following expression.
w _ji = tf _ji × idf _j

The feature vector D _j ′ of the document D _j is expressed as D _j ′ = (w _j1 , w _j2 ,..., W _jm ). Here, m is the number of dimensions in the vector space, that is, the total number of terms appearing in the document set. Then, 'the, D _q' feature vector D _q document _{D q = (w q1, w} q2, ···, w qm) expressed as a document D _j, the similarity sim between D _q (D _j, D _q ) can be calculated by the following equation.
sim (D _j , D _q ) = D _j ′ · D _q ′ / (| D _j ′ || D _q ′ |) = cos θ
= (W _q1 w _j1 + w _q2 w _j2 +... + W _qm w _jm ) /
(√ (w _q1 ² + w _q2 ² +... + W _qm ² ) (√ (w _i1 ² + w _i2 ² +... + W _im ² )

  The concept of “UNIX (registered trademark)” in “infoseek” and “YAHOO! When the degree of approximation S (UNIX (registered trademark), Unix (registered trademark)) with the concept of “Unix (registered trademark)” is calculated, as shown in FIG. With respect to the concept of “registered trademark”, the degree of approximation is calculated for all concepts under the UNIX (registered trademark) category of the Internet directory “YAHOO!”.

  And the concept of “UNIX (registered trademark)” of “infoseek” and “YAHOO! When the approximations are calculated for all the concepts under the UNIX (registered trademark) category of the Internet directory of "", the "YAHOO" for all the concepts under the UNIX (registered trademark) category of the "infoseek" Internet directory are calculated. ! "Is included in the Internet directories of" infoseek "and" YAHOO! "In FIG. 5 by calculating the approximate degree of brute force with all concepts under the UNIX (registered trademark) category of the Internet directory of"! " Mapping between concepts.

FIG. 8 is a diagram illustrating an example of a mapping result between the concepts of FIG.
In FIG. 8, as a result of mapping between the concepts included in the “infoseek” and “YAHOO!” Internet directories in FIG. 5, the “UNIX (registered trademark)” concept of “infoseek” and the “Unix” of “YAHOO!” The concept “FreeBSD” in “infoseek” and the concept “BSD” in “YAHOO!” Are mapped, and the concept “Solaris” in “infoseek” and “YAHOO!” It is assumed that the concept of “Solaris” is mapped.

Also, the “Linux” concept of “infoseek” is mapped to both the “UNIX (registered trademark)” concept of “YAHOO!” And the “Linux” concept of “YAHOO!”.
Furthermore, the concept of “newspaper and magazine” in “infoseek” and the concept of “support and documents” in “YAHOO!” Are mapped, and the concept of “Turboline” in “infoseek” and “TurboLinux” in “YAHOO!” It is assumed that the concept is mapped.
Then, when mapping is performed between the concepts included in the “infoseek” and “YAHOO!” Internet directories in FIG. 5, the mapping result of the parent-child class is checked for these mapped Internet directory classes. Thus, topology mapping between the concepts in FIG. 5 is performed.

FIG. 9 is a diagram illustrating an example of a topology mapping result between the concepts of FIG.
In FIG. 9, referring to the mapping result of FIG. 8, for example, the concept “Solaris” of “infoseek” and the concept of “Solaris” of “YAHOO!” Are mapped to each other. Then, when checking the mapping state between the parent concept “UNIX (registered trademark)” of the concept “Solaris” of “infoseek” and the parent concept “Unix (registered trademark)” of the concept “Solaris” of “YAHOO!” From the mapping result of FIG. 8, it can be seen that the concept “UNIX (registered trademark)” of “infoseek” and the concept “UNIX (registered trademark)” of “YAHOO!” Are mapped to each other.

As a result, it is determined that the concept set {UNIX (registered trademark), Solaris} in “infoseek” and the concept set {Unix (registered trademark), Solaris} in “YAHOO!” Can form a topology mapping.
Then, the above operation is executed for all the mapped concepts under the UNIX (registered trademark) category of the Internet directory of “infoseek” and “YAHOO!”.

  As a result, it consists of a subdirectory α consisting of a concept set {UNIX (registered trademark), Solaris, Linux, FreeBSD, newspaper and magazine} and a concept set {Unix (registered trademark), Solaris, Linux, BSD, support and document}. A topology mapping is formed with the subdirectory α ′. Further, a topology mapping is formed between the subdirectory β consisting of the concept set {Linux} and the subdirectory β ′ consisting of the concept set {Unix (registered trademark)}. Further, a topology mapping is formed between the subdirectory γ composed of the concept set {Turboline} and the subdirectory γ ′ composed of the concept set {TurboLinux}.

If a multiplexed topology mapping exists as a result of the topology mapping, a subdirectory that provides the multiplexed topology mapping result is detected. When a subdirectory that gives a multiplexed topology mapping result is detected, a topology mapping result formed from subdirectories other than the subdirectory having the largest number of concepts constituting the subdirectory is detected as an error.
That is, referring to the topology mapping result of FIG. 9, the subdirectories α and α ′ and the subdirectories β and β ′ are detected as subdirectories that give the multiplexed topology mapping result.

  When the subdirectories α and α ′ and the subdirectories β and β ′ that give the multiplexed topology mapping result are detected, the number of concepts included in the subdirectories α and α ′ and the subdirectories β and β The number of concepts included in ′ is compared. Here, the subdirectory α includes the concepts “UNIX (registered trademark)”, “Solaris”, “Linux”, “FreeBSD”, “newspaper and magazine”, and the subdirectory α ′ includes “Unix ( (Registered trademark) ”,“ Solaris ”,“ Linux ”,“ BSD ”,“ support and documents ”. Further, the subdirectory β includes the concept of “Linux”, and the subdirectory β ′ includes the concept of “Unix (registered trademark)”. Therefore, when the number of concepts included in the subdirectories α and α ′ is compared with the number of concepts included in the subdirectories β and β ′, the number of concepts included in the subdirectories α and α ′ is larger than the number of concepts included in the subdirectories α and α ′. It can be determined that the number of concepts included in the directories β and β ′ is smaller.

When it is determined that the number of concepts included in the subdirectories β and β ′ is smaller than the number of concepts included in the subdirectories α and α ′, as shown in FIG. The topology mapping result between β ′ is determined to be an error.
When an error in the topology mapping result between the subdirectories β and β ′ is detected, the mapping result between the concepts is corrected so as to follow the topology mapping result in which the error is corrected.

FIG. 11 is a diagram illustrating an example of the mapping result after removing the mapping error between the concepts of FIG.
In FIG. 11, as a result of the mapping of FIG. 8, the concept of “Linux” of “infoseek” is the concept of “Unix (registered trademark)” of “YAHOO!” And the concept of “Linux” of “YAHOO!”. Mapping with both. Then, by removing the mapping result between the concept of “Linux” in “infoseek” in FIG. 8 and the concept of “UNIX (registered trademark)” in “YAHOO!”, The subdirectories β and β in FIG. An error in the topology mapping result between 'can be corrected. As a result, it is possible to output a result obtained by mapping the concept of “Linux” of “infoseek” only with the concept of “Linux” of “YAHOO!”.

  The present invention can automatically perform mapping between concepts included in different information sources in consideration of the hierarchical structure of concepts, and can be used for a search engine that searches an instance of an Internet directory. it can.

It is a block diagram which shows schematic structure of the information management apparatus which concerns on one Embodiment of this invention. It is a figure which shows an example of the structure of the information source which concerns on one Embodiment of this invention. It is a flowchart which shows the information management method which concerns on one Embodiment of this invention. It is a figure which shows the topology mapping extraction method which concerns on one Embodiment of this invention. It is a figure which shows the structural example of the Internet directory of "infoseek" and "YAHOO!". FIG. 6 is a diagram illustrating an example of a method for measuring the degree of approximation between concepts included in each of the information sources 21a and 21b in FIG. It is a figure which shows an example of the approximation measuring method by the brute force method between the concepts of FIG. It is a figure which shows an example of the mapping result between the concepts of FIG. It is a figure which shows an example of the topology mapping result between the concepts of FIG. It is a figure which shows an example of the removal result of the topology mapping error between the concepts of FIG. It is a figure which shows an example of the mapping result after removing the mapping error between the concepts of FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1a, 1b, 21a, 21b Information source 2 Inter-concept mapping extraction apparatus 2a Approximation degree measurement means 3 Topology mapping extraction apparatus 3a Mapping state determination means 3b Connection form determination means 4 Topology mapping evaluation apparatus 4a Partial information source detection means 4b Concept number calculation It means 4c topology mapping error detection unit 5 mapping results _{11, C a1 ~C a3, C} b1 ~C b3 concept 12 instance O _A, O _B ontology _{O SA1, O SB1, O SA2} , O SB2 sub ontology

Claims (6)

Inter-concept mapping extraction means for extracting concepts that are similar to each other included in each of the two information sources;
Based on the topology of similar concepts extracted by the inter-concept mapping means, partial information sources that are approximated between partial information sources composed of concept sets that are in parent-child relationship are extracted to form topology mapping A topology mapping extraction means,
A partial information source in which the topology mapping result extracted by the topology mapping extraction means is multiplexed is detected, and an error in the topology mapping result is detected based on the number of concepts included in the detected partial information source. An information management apparatus comprising: topology mapping evaluation means for correcting a mapping result between concepts extracted by the inter-concept mapping extraction means by removing a topology mapping result in which an error is detected. The inter-concept mapping extraction means includes:
The information management apparatus according to claim 1, further comprising: a degree-of-approximation measuring unit that measures the closeness between instances of the concepts included in the two information sources. The topology mapping extraction means includes
Mapping state determination means for determining a mapping state of a concept having a parent-child relationship with the mapped concept;
Connection form determining means for determining the closeness of the connection forms of the concepts that are similar to each other extracted by the inter-concept mapping extraction means,
The topology mapping extraction unit forms a topology mapping between partial information sources composed of concept sets having a parent-child relationship with each other based on the proximity of the connection mode determined by the connection mode determination unit. The information management device according to claim 1 or 2. Mapping between concepts for two sources;
Determining the closeness of the connection form between concepts that approximate each other based on the parent-child relationship between the mapped concepts;
Based on the approximation of the connection form between the concepts, extracting a partial information source approximated between partial information sources composed of concept sets approximated by the connection form, and performing topology mapping;
Detecting a partial information source in which the topology mapping result is multiplexed;
Calculating the number of concepts included in the detected partial information source;
Detecting a topology mapping result error based on the number of concepts included in the partial information source, and correcting the mapping result between the concepts by removing the topology mapping result in which the error is detected. An information management program that is executed by a computer. Detecting an error in the topology mapping result,
  Causing the computer to execute a step of detecting, as an error, a topology mapping result formed from partial information sources other than the partial information source having the largest number of concepts constituting the partial information source among the detected partial information sources. 5. The information management program according to claim 4, wherein The step of performing topology mapping based on the approximation of the connection form between the concepts includes:
  Checking the mapping result of the parent concepts of the concepts mapped to each other;
  If the parent concept is mapped, forming a topology mapping between partial information sources consisting of concepts mapped to each other and a concept set including the parent concept;
  Confirming a mapping result of a concept child concept included in the partial information source in which the topology mapping is formed;
  When the child concept is mapped, a computer is caused to execute topology mapping between partial information sources including concepts mapped to each other, a parent concept thereof, and a concept set including the child concept. The information management program according to claim 4 or 5.

Images