Research on the Method of Answer Extraction in Domain Chinese Question Answering System

The domain knowledge has a direct influence on the result of question-and-answer in the specific domain question answering system.In this paper,a method of answer extraction from domain question-and-answer is presented,in which candidate paragraphs can be obtained for answer extraction with the help of domain knowledge retrieval based on question query,question types and answer types.For definition questions,sentences or paragraphs with higher relevance can be extracted to become the answer based on the relevance ranking between questions and the candidate sentences or paragraphs by combining the computing method of keywords weighting with the method of semantic similarity between the sentences and questions.For questions of numeral or list entity type,with the help of named entity recognition,the domain entity related to the key part of the questions is extracted to be the answer.The experiments on the answer extraction have been carried out in the field of Yunnan tourism and the results show that favorable effects have been achieved by using the method of answer extraction from domain question-and-answer.