Deep question answering for single documents with lexical chains

作者：Cao Ling; Qiu Xipeng; Huang Xuanjing^*

来源：Journal of Computational Information Systems, 2011, 7(16): 5859-5865.

• Answer extraction
• Coreference
• Data sets
• Information redundancies
• Lexical Chain
• Machine reading
• Name entity recognition
• NAtural language processing
• Open domain question answering
• PoS tagging
• QA system
• Question analysis
• Question Answering
• Reading comprehension
• Semantic analysis
• Semantic relations

The pipeline of traditional open domain question answering (QA) is question analysis, retrieval, answer extraction, answer selection and validation. Most QA systems can achieve better performances with large corpora, such as Web, by taking advantage of information redundancy, which can simplify the jobs of the other parts. In this paper, we focus on deep question answering for single documents, in which semantic analysis is a more important question analysis. We proposed a method based on lexical chains [10], which works in the same way people do in reading comprehension tests. Lexical chains are used to estimate the semantic relation between texts. Moreover, natural language processing (NLP) techniques are also widely used in our method, such as: POS tagging, name entity recognition, coreference. Experiments show our method is efficient on the QA4MRE dataset. ? 2011 by Binary Information Press.

• 出版日期2011
• 单位复旦大学