基于分词的语句相似度计算的改进

摘要

Sentence similarity reflects the similarity degree between two sentences. Similarity computing is one of the core technologies of the FAQ and information retrieval. This paper, in view of the of over-reliance on real word effect in sentence similarity calculation based on word segmentation, the word form similarity calculation without word segmentation is added in the original calculation model, in order to alleviate the low level of similarity calculation result due to the inaccurate sentence segmentation. Experiment results obtained by combining with "Data Structure" Question-Answering System show that the improved method has a better accuracy than the original.%语句相似度体现的是两个句子之间的相似程度.语句相似度计算是FAQ和信息检索等方面核心技术之一.针对基于分词的相似度计算过于依赖实际的分词效果,在原相似度计算模型中增加了两个句子不分词时的词形相似度计算,以缓解因为句子分词不准确而导致相似度计算结果偏低的情况.结合“数据结构”课程问答系统的实验,结果表明,改进的方法比原方法有较高的准确率.