Extracting Hyponymy of Ontology Concepts from Patent Documents

Automatic extraction of hyponymy relations between concepts in an ontology is significant for ontology learning and knowledge organization. In this paper, we propose a fusion approach of hyponymy relation extraction in patent domain, using Relative Decoration Degree (RDEG) to extract high precision relations, and then Association Rule (AR) to enrich those relations. We use Cilin to extend a word to a set to improve the recall, and we consider the position of core words in Chinese to improve the precision. Different to classical studies, we use a simplified method to select parameters to fuse approaches and merge extraction relations together as the final result. The results comparing with the baseline show that this approach can obtain better performance on the hyponymy relation extraction task.