Automatic Extraction Of Semantic Concept-Relation Triple Pattern From Wikipedia Articles

In the existing methods for building and extension of Ontology, there is a manually worked form by professionals or a semi-automated method that uses the probability distribution of statistics through analysis of universal dictionary or thesaurus group. If it is produced manually, its accuracy for concept extraction and relation production is excellent, but it requires a lot of time and money. To solve these problems in the semi-automated method, there are differences in the interpretation of words tagged when analyzing the text and it relies on the universal dictionaries or learning articles for concept and relation extraction. This has a disadvantage that the Ontology building and extension are limited before referenced article is modified. For this, this paper analyzed the link pattern of Link Grammar after extracting the terminology within the Wikipedia articles which represent the collective intelligence, and proposed the domain Ontology extension method that extracts the Triple pattern describing the relation between concepts. The order has been determined by assigning the weight for each extracted relation and concept through the proposed method. For the results of this test, after extracting concept-relation of 5,100 key sentences extracted from the Wikipedia articles using Link Grammar, the results were evaluated.