融合情感词典与深度学习的文本情感分析研究

Research on Text Sentiment Analysis Based on Sentiment Dictionary and Deep Learning

文本情感分析是自然语言处理领域内的重点研究方向。当前Word2vec词向量结合神经网络的深度学习模型在中文文本情感分析研究中取得了不错的成绩。然而仅使用词向量模型作为文本表示进行模型学习时,会丢失当前词的情感信息。故论文提出一种基于情感词典结合双向长短期记忆网络和注意力机制的文本情感分析模型SABLSTM。该模型在酒店数据集上的分类准确率是93.17%,比仅结合了注意力机制的双向长短期记忆网络模型的准确率提升了1.56%。由此可见,以情感词典作为先验知识进行模型训练,可以提升中文文本情感分析任务的效果。

The sentiment analysis is a crucial part of natural language processing.At present,the deep learning model of word embedding combined with neural network has achieved good results in the research of Chinese text sentiment analysis.However,the emotional information of the current word will be lost when using word embedding model only as text representation to learn the mod-el.In this work,a text sentiment analysis model SABLSTM is proposed,which is based on sentiment dictionary combined with bidi-rectional long short-term memory network and attention mechanism.The precision of this model on the hotel dataset is 93.17%,and it has achieved a great improvement of 1.56%compared with the model integrated the bidirectional long short-term memory network structure with attention mechanism.Therefore,the model training with sentient lexicon as prior knowledge can improve the effect of Chinese text sentiment analysis task.