基于改进双向LSTM的评教文本情感分析

Sentiment analysis of teaching evaluation text based on improved bidirectional LSTM

为提高教评文本情感分析的准确率和适用性,提出一种利用改进Bi-LSTM结合Word2Vec的评教文本情感分析方法。采用词向量(Word2Vec)方式将每个文本转换为输入层中特征空间的句子表达,完成词向量的转换;改进Bi-LSTM以解决深度神经网络的退化问题,结合语法规则,构建情感分类器。基于Python开发环境对所提方法进行实验对比,实验结果表明,所提方法的分析性能优于其它对比方法,其准确率、精度、召回率和F1值分别达到93%、89%、97%和92%,具有较理想的分析能力。

To improve the accuracy and applicability of sentiment analysis of teaching evaluation text,a sentiment analysis method of teaching evaluation text based on improved Bi-LSTM and Word2vec word vector was proposed.The word vector(Word2Vec)method was used to transform each text into sentence expression of feature space in the input layer.Bi-LSTM was improved to solve the degradation problem of deep neural network.Combined with the grammar rules,the sentiment classifier was constructed.Experimental results based on Python development environment show that the performance of the proposed method is better than that of other comparison methods.The accuracy,precision,recall and F1 value of the proposed method are 93%,89%,97%and 92%,respectively.The proposed method has ideal analysis ability.