A Pattern-Based Approach to Semantic Relation Extraction Using a Seed Ontology

Abstract

This paper presents our experiment on the "Badea" system. A system designed for the automated extraction of semantic relations from text using a seed ontology and a pattern based approach. We describe the experiment using a set of Arabic language corpora for extracting the antonym semantic relation. Antonyms from the seed ontology are used to extract patterns from the corpora, these patterns are then used to discover new antonym pairs, thus enriching the ontology. Evaluation results show that the system was successful in enriching the ontology with over 400% increase in size. The results also showed that only 2.7% of the patterns were useful in extracting new antonyms, and thus recommendations for pattern scoring are presented in this paper.