A statistical approach for semantic relation extraction

Semantic relations are an important component of ontologies that can support many applications e.g. text mining, question answering, and information extraction. Automatic semantic relation extraction system is a crucial tool that can reduce the bottleneck of knowledge acquisition in the ontologies construction. In this paper, we present a statistical approach for learning the semantic relations between concepts of an ontology in the agricultural domain. The semantic relations are acquired by using verbs to indicate the relations between ontology concepts. The co-occurrences of domain-verbs with their components, which are annotated the concepts, are analyzed by using several statistical methodologies. Moreover, we expand the sets of verb expressing the same semantic relation by using the extracted patterns of concept pairs of the seed verb's component. Our experiment has been done on a collection of Thai shallow parsed texts in the domain of agriculture. The precision and recall of the presented system is 65% and 82%, respectively.