KR20110026039A - Ontology matching method using broader terms - Google Patents

Abstract

PURPOSE: An ontology matching method using broader terms is provided to supply a result most proper to the expansion of ontology by deducing a category system relationship that is most proper to the category system expansion of given ontology. CONSTITUTION: The broader terms for given words are extracted by using a category structure of structured data(S10). An attributing graph is deduced from the broader terms(S20) and is analyzed by using a hits link analysis algorithm(S30). The matching score for each concept of ontology is generated by using the analysis result(S40). The concept of the ontology matched with a given word is selected based on the matching score(S50).

Description

Ontology matching method using broader terms}

The present invention relates to a method of matching an input word to one of the concepts of ontology by analyzing the broad terms, given a word, broad terms and an ontology of the word. Construct a mathematical graph using words, analyze the mathematical graph using a link analysis algorithm, and use the analyzed results to score matching concepts on the ontology, and then score the concept to match words. It relates to a matching method for selecting based on. Words that match a concept become instances or subclasses of the concept, and this process can extend the taxonomy of a given ontology.

In case the user cannot input a broad term, this algorithm attempts to obtain broad terms using Wikipedia's category structure.

Classification systems play a very important role in many types of intelligent systems. However, building a classification system manually requires a lot of time and effort. Therefore, many researchers have been studying how to build the classification system automatically.

In order to build a taxonomy automatically, we must first obtain information that will be used to construct the taxonomy. In general, retrieval of taxonomy information uses two resources. The first is to get information from a general unstructured document. Since Hearst defined four lexico-syntactic patterns to extract taxonomy relationships in his 1992 paper, "Automatic Acquisition of Hyponyms from Large Text Corpora," most subsequent studies use this pattern, Or it has progressed in the direction of expanding a pattern. The second is to get information from the structured data. Recently, due to the availability of a large amount of structured data such as Wikipedia's category structure, a method of obtaining information of the classification system from the structured data has been activated. Ponzetto's 2007 paper, "Deriving a Large Scale Taxonomy from Wikipedia," attempted to derive a large number of taxonomy information from Wikipedia's category information using a variety of methods.

The present invention proposes a method of extracting a taxonomy relationship for extending a taxonomy of an ontology existing from a word and its broad terms. This approach is directly applicable to Wikipedia's category structure, or to a number of existing structured data such as thesaurus. The existing method of extracting classification system relations using broad terms mainly determines whether a pair of <words, broad terms> is related to a classification system, or the method proposed in the present invention collects information from a set of broad terms. By extending, we can obtain a more accurate and taxonomy relationship that is appropriate for the taxonomy expansion of a given ontology.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems. The present invention takes an ontology, a word, and a broad term of the word as input, and expands the classification system of a given ontology using the word given as input. The goal is to obtain information about the most appropriate taxonomy relationship.

Extraction method of the classification system relationship using broad terms according to the present invention for achieving the above object,

Extracting broad terms of a given word using a Wikipedia category structure; A second step of deriving an Attributing Graph from the obtained broad terms; A third step of analyzing the equation graph using the HITS link analysis algorithm; A fourth step of assigning a matching score to each concept of the ontology using the analysis result of the equation graph; The fifth step is to select a concept of ontology to match a given word using the matching score.

The ontology concept matching method using broad terms according to the present invention as described above is best suited for ontology extension compared to other existing algorithms by deriving a classification system relation that is most suitable for extending the classification system of a given ontology. You can expect a system that will yield results.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. In describing the present invention, detailed descriptions of related well-known functions or configurations are omitted in order not to obscure the gist of the present invention.

1 is a flowchart illustrating a method of matching a given word to a concept of a target ontology through a broad term according to the present invention. Here, matching means a concept in which a given word is semantically included.

First, the system extracts the broad terms of a given word using the Wikipedia category structure (S10). This is done by simply looking for a Wikipedia page or category name with the same name as the given word and getting the top three category names. This process can be replaced by user input.

Next, a mathematical graph is constructed using a set of broad terms that have passed through step S10 or obtained through user input (S20). FIG. 2 shows an input example of the system, and FIG. 3 shows a mathematical graph constructed from this input. In order to construct a mathematical graph, we first extract the head words of each broad term, using the method presented in M. Collins' 1999 paper, "Head-driven Statistical models for natural language parsing." As shown in Fig. 3, each node of the constructed equation graph represents each token extracted from the broad term. The sides of the equation graph indicate the existence and frequency of the mathematical relationship between the modifiers and each head word. In this case, it is assumed that the mathematical relationship is that if a head word and a non-head word exist within the same broad term, the non-head word always modifies the head word. For example, in FIG. 3, the meaning of the value of 4 from 'by' to 'Albums' is the meaning of the words 'Albums' as the head word among the broad terms of the given word !! Destroy-Oh-Boy !! In other words, there are four occurrences of 'by'.

In the equation graph generated through the step S20, the 'HITS page ranking algorithm' presented in J. M. Kleinberg's 1999 paper "Authoritative sources in a hyper-linked environment" is used to score each node (S30). The HITS algorithm scores two scores: the HITS authority score and the HITS hub score. The HITS authority score determines how many pages are referred to by other pages. The 'HITS hub score' indicates how many pages (nodes) 'reference' other pages.

After each node is scored by the step S30, a score is scored for each concept of the ontology given by the formula of FIG. 4 (S40). This score is the parent of the taxonomy of a given word, indicating how well each concept of the ontology fits. In the formula of FIG. 4, c represents a target concept, h represents a head word of a word representing a target concept, and A represents a set of tokens excluding a modifier from a word representing a target concept.

When the score for each concept of the ontology is scored by the step S40, find the highest scored concept and attach the given word as 'subclass' or 'instance' of the concept (S50). In the present invention, the emphasis has been placed on extending the classification system, so it is clear that no special classification is made between subclasses and instances. If all concepts score 0, the expansion of the ontology through the given word is treated as a failure.

As described above with reference to the drawings illustrating a method of matching a given word to one of the concepts of a given ontology using a broad term according to the present invention, the present invention is limited by the embodiments and drawings disclosed herein Of course, various modifications can be made by those skilled in the art within the technical scope of the present invention.

1 is a flow chart of a method for matching a given word to one of the concepts of a given ontology, using the synonyms of that word, in accordance with the present invention.

2 is an illustration for showing the implementation of the present invention.

Figure 3 is a formula graph that is constructed when the input of Figure 2 in the present invention.

4 is a formula for scoring the concept of each ontology using the analysis result of the equation graph in the present invention.

Claims (6)

Ontology matching method of words through broad terms including the following steps: Extracting broad terms of a given word using a category structure of structured data; A second step of deriving an Attributing Graph from the broad terms obtained in the first step;  A third step of analyzing the equation graph obtained in the second step by using an HITS link analysis algorithm; A fourth step of assigning a matching score to each concept of the ontology using the analysis result of the equation graph obtained in the third step; And A fifth step of selecting a concept of ontology to be matched to a given word by using the matching score obtained in the fourth step. The method of claim 1, wherein each node in the equation graph of the second step represents a respective token extracted from the broad term, and the sides represent a broad term such that the existence and frequency of the mathematical relationship between the modifiers and each head word are represented. Method of ontology mapping of words through The method of claim 1, wherein each node in the third step is scored using a HITS algorithm, which differs from the HITS authority indicative of how many nodes (pages) are referenced by other pages. A method of ontology mapping of words through a broad term, with a HITS hub score that indicates how many pages are modified. The method of claim 1, wherein the method for scoring each concept of the ontology of the fourth step is a method of ontology mapping of words through a broad term using the following equation (where c is an object concept and h is a word representing an object concept). head word, and A represent a set of tokens excluding modifiers in words representing target concepts)

The ontology mapping method of claim 1, wherein the structured data of the first step uses a web document including Wikipedia. Ontology mapping system of words through broad terms, consisting of the following components: A first processor extracting the broad terms of a given word using a web document category structure; A second processing unit for deriving a mathematical graph from the broad terms obtained by the first processing unit; A third processor for analyzing the equation graph obtained by the second processor using an HITS link analysis algorithm; A fourth processing unit for assigning a matching score to each concept of the ontology using the analysis result of the equation graph obtained by the third processing unit; And A fifth processor selecting a concept of ontology to be matched to a given word by using the matching score obtained by the fourth processor.

Images