Automatic ontology development from semi-structured data in web-portal: Towards ontology of thai rice knowledge

Abstract

Heavyweight ontology is difficult to develop even for experienced ontology engineer, but it is required for semantic based computer software as core knowledge. Most of existing automated ontology development methods however focuses on lightweight ontology, taxonomy-instance extraction. This work presents a method to automatically construct relation-heavy ontology from semi-structured web content providing deep knowledge in specific domain. Classes, instances and hierarchical relation are derived from the category content from the web. Relations are extracted based on frequent expression details. Templates of relation and its range are extracted from common content with partial difference. Similar contexts are grouped with similarity and form as relation to attach to ontology classes. The case study of this work is Thai rice knowledge including rice variety, disease, weed and pest provided in website from responsible government. The complete ontology is used as core knowledge for personalised web service. The service assists in filter content in summary that matched to users’ information. Courtesy to the generated relation-heavy ontology, it is able to recommend relevant chained concepts to users based on semantic relation. From evaluation from an expert, the generated ontology obtained about 97% accuracy.