基于TextCNN的文本情感分类系统

Text sentiment classification system based on TextCNN

通过分析用户在线评论的文本信息来预测消费者的网购偏好意愿,进而提高消费者的满意度成为众多企业的需求.但庞大的评论数据量使得人工手动对评论文本进行分类打标签难以实现,结合Word2vec和TextCNN模型实现对在线评论进行文本情感分类.对评论文本进行规格化处理,通过结巴分词库等对已处理数据进行分词,即提取关键字词.使用Word2vec工具对每个分词进行词向量的训练,得到word embedding权重矩阵作CNN模型的嵌入层,采用TextCNN模型训练得到本文的情感分类模型.相比于直接用传统的卷积神经网络CNN默认的词嵌入层,本文训练出来的神经网络模型效果更佳.

It had become the demand of many companies to predict consumers′ online shopping preferences by analyzing the text information of users’ online comments, and thereby improved consumer satisfaction. However, the huge amount of comment data made it difficult to manually classify and label the comment text. This paper combined the Word2 vec and TextCNN models to implement text sentiment classification for online comments. The comment text was normalized, and the processed data was segmented through the word segmentation database, etc., that was, the keyword words were extracted. Used the Word2 vec tool to train the word vector for each word segmentation, got the word embedding weight matrix as the embedding layer of the CNN model, used the TextCNN model to train to get the sentiment classification model of this paper. Compared to directly using the default word embedding layer of the traditional convolutional neural network CNN, the neural network model trained in this paper had a better effect.