Semantic Querying based Concept Hierarchy Construction for Ontology Learning
B. Sathiya, T. V. Geetha · 2016
The method of identifying a set of concepts (domain specific) and the relations among these concepts from text is called ontology learning. These relations can be taxonomical (hypernym and hyponym) or non-taxonomical relations. The important step in ontology learning is to, extract the concepts representing the domain and building the concept hierarchy based on the taxonomical relations prevailing between them. The important resources of the text used for concept hierarchy construction are domain-specific corpus and vast text from web pages. The former resource is static and most likely outdated information, whereas the latter resource is uncertain. Therefore, to tackle these drawbacks, we have proposed a new two-level semantic query formation methodology which is based on lexical-syntactic patterns. It utilizes both the resources of text: corpus and web pages for automatic concept hierarchy construction by discovering the hypernyms and hyponyms. Specifically, it resolves the limited and static content problem of the corpus by utilizing the vast and current knowledge available from the web. Meanwhile, the uncertainty of the knowledge in the web is removed by adding two new contextual information to the semantic queries. From the experimental results, it is evident that the proposed concept hierarchy construction method achieved enhancement of 6.5%, 5.65% and 6.8% for metrics such as Precision, Recall, and F-Measure respectively.