一种基于语义与结构的句子相似度计算方法

A Method of Sentence Similarity Computing Based on Semantic and Structure

在对已有方法进行分析的基础上,提出语义与结构相结合计算句子相似度的方法,并系统地介绍了此方法的实现过程。首先对输入的句子进行预处理,得到词语序列,使用基于知网的方法得到词语间的语义相似度;然后使用词形和句长特征表示句子的结构相似度;最后加权得到两个句子的相似度。实例证明,提出的方法计算得到的句子相似度取得了较好的效果。

Sentence similarity reflects the text similarity, and similarity computation occupies a very important role in the field of natural language processing. After analysing the existing methods, a method based on semantic and structure to calculate sentence similarity was presented, of which the implementation was introduced systematically. First, the input sentences were preprocessed to obtain words lists, while the approach based on hownet was used to get semantic similarity between words. And then, the word form and structural feature were described as sentence structural similarityl Finally, similarity between two sentences was obtained through weight calculation. Experimental examples show that the method proposed in this paper to calculate sentence similarity is effectiveness.