Paragraph act based pragmatic information extraction in question answering
Paragraph act based pragmatic information extraction in question answering
复制标题
DOI：
10.1109/ccis.2011.6045051
复制
发表时间：
2011-10
期刊：
影响因子：
--
通讯作者：
Song Liu;F. Ren
中科院分区：
文献类型：
--
作者：
Song Liu;F. Ren
在问答系统中，最大的困难是跨越问题和语料库之间的差距来找到合适的答案。问题和语料库处于不同的上下文中，为了找到答案，需要推断一个候选句子是否符合用户的请求。实用主义关注的是对话语和用户的推断。它是跨越问题和语料库鸿沟的桥梁。本文分析了语用要素，并基于段落行为从语料库中提取语用信息。实用信息也组织得很好。实验表明，附加的语用信息显着提高了问答系统的性能，无论是查准率、查全率还是F分，尤其是在答案数量受到限制的情况下。
In Question Answering System the greatest hardship is crossing the gap between the questions and corpus to find the appropriate answers. The questions and corpus are in different context, to find the answers the inference whether one candidate sentence fits the user's request is necessary. Pragmatic is concerned with inference about the utterances and user. It is a bridge across the gap of questions and corpus. In this paper we analyze the elements of pragmatic and extract the pragmatic information from the corpus based on the paragraph act. The pragmatic information is also well organized. And the experiments show that the appended pragmatic information significantly improves the performance of the Question Answering System either the precision, recall or F score especially in the condition that the quantity of answer is restricted.