Anaphora resolution in Chinese texts based on domain ontology

This paper presents a framework 4 anaphora resolution including coreference resolution based on domain ontology. To acquire better features and mine semantic relationships of entities in documents, we develop domain ontology and make use of ontology knowledge rather than flat gazetteers, proper lexicons or term lists. The basic steps that we put forward are: creating domain ontology seeds; training resolution model by adopting adaptive Machine Learning algorithms; gaining new knowledge and recognizing anaphora relations. Then in section 5, we put forward the strategy of semantic similarity computing and expect to mine semantic similarity feature in specific domain with general linguistic ontology resource-HowNet(1), so as to improve the performance of NPs' coreference resolution. The concrete method is to acquire NPs' concept similarity from sememe similarity.