融合语义依存关联对的句子相似度的负面在线评论案例推理

及时有效地处理负面在线评论能提高企业的声誉和维护消费者对企业的信任,而面对大量的负面在线评论,企业很难在短期内及时有效回复.根据在线评论是短文本的特点,运用基于句子相似度的负面在线评论进行案例推理处理.针对负面在线评论句子相似度计算问题,抽取句子中评价对象词和评价词作为关键词,同时考虑评价词的修饰词和句子语义依存关联对.除提出的关键词、修饰词、语义依存分析关联对三个特征外,再融入句子中的公共词、语义等两个特征,提出一种融合公共词、关键词、修饰词、语义、语义依存分析关联对等五个特征的句子相似度计算方法,检索相似负面评论句子的案例.通过实验验证,融入多特征的句子相似度计算优于常见句子相似度计算方法,其准确率为83.3％,有效地检索案例句子集并自动推荐回复方案.

Negative online review case-based reasoning on fusing semantic dependency association pairs of sentence similarity

Timely and effective treatment of negative online reviews can improve the reputation of enterprises and maintain the trust of consumers. However, facing a large number of negative online reviews, it is difficult for enterprises to respond effectively in a short time. According to the characteristics that online reviews are short texts, negative online reviews based on sentence similarity are used for case-based reasoning. Aiming at the problem of sentence similarity calculation of negative online reviews, this paper extracted the evaluation object words and evaluation words as keywords, and considered the modifiers of evaluation words and sentence semantic dependency association pairs. In addition to the three features of keywords, modifiers and semantic dependency analysis association pairs, this paper proposed a sentence similarity calculation method which integrated five features of common words, keywords, modifiers, semantics and semantic dependency analysis association equivalence to retrieve similar negative reviews sentences. The experimental results show that the sentence similarity calculation method based on multi features is better than the common sentence similarity calculation methods, and its accuracy rate is 83.3%. It can effectively retrieve the case sentence set and automatically recommend the reply scheme.