基于位置信息多注意交互模型的方面级情感分析

Multi-attention Interactive Model with Position-information for Aspect Level Sentiment Analysis

方面级情感分析是用于识别文本中特定方面情感倾向的细粒度级情感分析。针对传统神经网络模型在上下文和方面词之间交互性较少,不同距离上下文对方面词情感极性判断有不同影响等问题,提出了基于位置信息多注意交互模型(MAIP)。该模型引入依存句法树表示词语之间的依存关系,将位置信息嵌入到词向量中,使用双向GRU获取方面词和上下文的隐藏向量表示,并利用多注意力机制学习词语之间的交互信息。实验结果表明,该模型能有效提升情感分析的准确度。

The purpose of aspect-level sentiment analysis is to identify the emotional tendency of a particular aspect in a text.It is a fine-grained sentiment analysis.To solve the problems of the traditional attention-based neural network model,such as the less interaction between context words and target words,and the different contributions of context words from different distance to the judgment of emotional polarity of aspect words,a multi-attention interactive model with position-information is proposed.This model introduces the dependency syntax tree to represent the dependency relationship between words,inserts the position information into the word vector,the hidden vector representation of aspect and context is obtained by BIGRU,and then the important feature information between context and aspect is learned by using the multi-attention interactive model.The experimental results show that the model can effectively improve the accuracy of emotion analysis.