Application of NER and MC in Answers Extraction of Factoid Questions

Question answering (QA) system is intended to be fast, simple and accurate to give feedback to the user's natural language questions. By studying answers extraction in the Chinese QA system, this paper adopted the method of combination of the Named Entity Recognition (NER) and the theory of Metric Cluster (MC) to extract answers about factoid questions. These answers extraction experiments used the standard of TREC evaluation, and the factoid questions' average MRR value was 0.6883. These results show that the method had a higher MRR value and good effect on answers extraction.