结合句子序列与语法关系的方面级情感分类方法

An aspect-level sentiment classification method combining sequence and syntax

为了研究句子结构关系对方面级情感分类模型性能的影响,针对当前方面级情感分类方法中过于重视句子中语法关系的作用,而忽视了句子的序列结构关系的问题,提出一种结合句子序列与语法关系的信息融合网络(sequence-syntax information fusion network,SYFN)模型。SYFN模型分别处理句子的序列与语法关系信息,并在融合2种结构关系信息的基础上进一步结合网络的高低层融合信息。这些改进措施使模型能够结合高低层网络的关系融合信息去处理各种复杂的句子关系,能够充分利用结构关系信息进行情感分类。实验结果表明,SYFN模型与基线模型相比性能有较明显的提升。

In order to study the influence of sentence structure relationship on the performance of aspect level sentiment classification model,a new information fusion network(sequence-syntax information fusion network,SYFN)model combining sequence and syntax relationships is proposed to solve the problem that the current aspect level sentiment classification methods pay too much attention to the role of grammatical relationship in sentences,while ignoring the sequence structure relationship of sentences.In the SYFN model,the sequence and grammatical relationship information of sentences are processed separately,and based on the fusion of the two structural relationship information,the high and low layer fusion information of the network is further combined.These improvements enable the model to combine the relationship fusion information of high and low layer networks to deal with various complex sentence relationships,and fully utilize structural relationship information for sentiment classification.The experimental results show that the performance of the SYFN model is significantly improved compared with the baseline models.