一种混合型的句子语义相似度计算方法

A Hybrid Measurement for Sentence Similarity Based on Semantic

句子相似度计算是智能答疑等应用领域的关键技术之一,当前相似度计算方法大都对语义因素考虑不全,使得计算的结果准确度差,为此提出了一种混合型句子语义相似度计算方法。首先,通过语义资源得到词间语义相似度;其次,以词之间的相似度为基础,结合词的权重和句型的相似度得到了混合的句子间的语义相似度,避免了单一角度度量相似度带来的偏差;最后,通过对比实验验证了本文提出的混合型计算方法相对传统方法获得了更好的测试结果。

Sentence similarity computation is one of the key technologies of intelligent question answering etc. The current method to calculate the similarity are not completely in terms of semantics, resulting the lack of accuracy. Therefore a hybrid type of sentence semantic similarity calculation method is proposed. First of all, word semantic similarity are gotten by semantic resources, and then the semantic similarity between sentences are gotten by mixedsimilarity with the word similarity and word weighting and sentence patterns, avoiding bias caused by single factor. Finally, the contrast experiment shows the proposed algorithm has better results compared with the traditional method.