Chinese definitional question answering based on soft pattern and ontology

A typical definitional question answering system extracts definition sentences from multiple documents and summarizes the sentences into definitions. In this paper, a new approach is proposed to extract definitional sentences from Web corpus, which is based on soft pattern and ontology. Soft pattern is used to judge whether sentences are definitional with syntactic analysis and ontology can deduce the definitional one with semantic information. The experimental result validates the effectiveness of this method, and the precision reaches 32.7% and is 5.2% higher than the method only based on soft pattern.