基于粒计算与语义模板的问答系统研究

项目名称： 基于粒计算与语义模板的问答系统研究

项目编号： No.61305094

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 自动化技术、计算机技术

项目作者： 徐菲菲

作者单位： 上海电力学院

项目金额： 25万元

中文摘要： 当前问答系统普遍存在问题分析不足、答案准确率低、智能性差等问题，在深入研究粒计算、语义模板的基础上，我们借助粒计算理论能发挥模拟人类思维从不同角度、不同层次分析问题的优势，采用语义和模式相结合的语义模板方法，实现基于粒计算与语义模板的问答系统，使得用户提出的问题能够被机器更深层识别和理解，从而提高问答系统的准确率与智能性。研究内容重点包括：针对目前模板技术并未考虑调整模板粒度使得问题分析准确率不高等问题，研究语义模板的定义及其粒化方法，以及模板粒度评估策略；针对目前问题分类的精度不高等问题，研究基于粗糙集的问题分类规则提取及相似问题判别方法；针对目前答案获取的准确率较低等问题，研究答案的粒度层次构建及选取标准，同一问题不同答案的聚类方法。本课题将粒计算理论与语义模板技术进行深层融合并应用到交互式问答系统中，为问答系统的研究提供一种新的有效解决方法，具有重要的理论研究与实际应用价值。

中文关键词： 用户交互式问答系统；粒计算；语义模板；粗糙集约简；

英文摘要： In order to improve the accuracy of question analyzing and answering thus to improve the intelligence of user-interactive question answering systems, we introduce granular computing into current question answering systems on the bais of our deep research on semantic pattern. Granular computing is an effective way to simulate human's thinking and problem solving abilities in terms of simplifying complex real world cases into multi-dimensional problems with multi-granularities. Enhanced withgranular computing and semantic patterns, the question answering systems can improve their understanding to user's questions, so that the accuracy of question answering and intelligence can be improved. This research mainly includes: since the existing pattern-based techniques do not consider pattern granularity for certain semantic requirement, we study the definition and the granulation of semantic patterns, and the method of granularity evaluation, to generate semantic patterns not only to meet the need of semantic analysis but also to cover more questions; To improve the accuracy of question classification, we address a question classification and rule extraction method based on rough sets, and the computational method of question similarity; To solve the problem of low quality of answers, we propose a hierarchical model o

英文关键词： User-Interacitve Question Answering；Granular Computing；Semantic Pattern；Rough Set Reduction；