基于主题相关性和深度学习的文本情感分析

Text Sentiment Analysis Based on Topic Correlation and Deep Learning

摘要: 将相关主题模型和神经网络相结合开展文本情感分析研究。首先, 为了度量文本的主题相关程度, 采用CTM模型对文本进行特征分割, 得到主题与词之间的相关矩阵和文本句子的主题特征向量; 其次, 基于相关性理论, 构造蕴含主题相关信息的词向量, 采用word2vec模型进行文本词表示; 最后, 使用BiLSTM模型对文本句子进行表示, 实现文本情感特征提取。

Abstract: The correlated topic model and neural network are integrated to carry out text sentiment analysis. Firstly, in order to measure the degree of topic correlation, the correlated topic model is used to segment the texts. Based on this, the correlation matrix and the topic feature vector of the sentences are obtained. Secondly, the word vectors containing topic related information are constructed based on the correlation theory, and the word2vec model is used to represent text words. Finally, BiLSTM model is used to represent text sentences and realize emotion feature extraction.