基于语义依存分析的句子相似性度量算法及应用研究

摘要

问答系统在MOOC(Massive Open Online Course)学习平台中占有很重要的地位,但由于MOOC课程学习中用户多、问题多的特点,快速准确地找到某一问题的答案,成为提升MOOC平台用户体验的一个挑战.为此,构建了面向常问问题集的问答系统,通过计算问句的相似性实现系统的自动问答.在现有的句子相似性度量方法基础上,设计了基于语义依存关系的度量方法,并结合了句长、词形、词义等多种特征度量句子之间的相似性.实验结果表明,该方法能够较好地反映句子之间的语义差别,是一种可行有效的方法,且提高了系统的响应准确率.该系统具有较高应答准确率,具有较强使用价值和广阔的应用前景.%Question answering system makes a big difference on MOOC learning platform, but since number of users and questions is so large that how to find the certain answer to a question quickly and accurately is a challenge to enhance user experience.So a frequently asked question system was designed to response question automatically.Based on the existing similarity calculation methods, this paper proposed a similarity calculation method according to semantic dependency parsing and multiple characteristics between sentences such as same words, sentence-length, and semantic of words.Test result shows that this method can reveal sentence difference better improve the accuracy of the response.This system has high response accuracy, has high practical value and wide application prospects.