Semantics-Based Answers Selection in Question Answering System*
W. Liu et al. (Eds.): ICWL 2004, LNCS 3143, pp. 354-362, 2004.，-0001，（）：
Motivated by the large requirement of Web answer in E-learning, this paper proposes a novel answer selection scheme of NL-WAS. With our scheme, the semantic types of the user' s question and the semantic templates were used to implement semantics-based answers selection. The semantic type is identified to reduce the retrieval scope firstly. When the semantic type can't be decided easily, the statistical similarity and the semantic similarity are computed to select the right answers. Otherwise, different strategies are proposed to match answers according to the type of user's question. If the semantic types between the user's question and one candidate' s are same, the same parts of their semantic templates are compared. When they have correlative semantic types, the corresponding parts of their semantic templates are compared according to the rules that are used to match different templates. The experimental results show that NL-WAS can answer the most test questions correctly.
版权说明：以下全部内容由郑庆华上传于 2005年12月22日 23时52分08秒，版权归本人所有。