基于注意力机制Bi-LSTM算法的双语文本情感分析

BILINGUAL TEXT SENTIMENT ANALYSIS BASED ON ATTENTION MECHANISM BI-LSTM

针对双语微博情感分析方法稀缺且准确率低的问题,根据相同英文词汇在不同语境下对文本情感作用不同这一事实,提出基于注意力机制的双语文本情感分析神经网络模型。该模型使用双向循环神经网络模型学习文本的特征表示,并引入注意力机制,为文本不同词语赋予不同权重,得到融合特征后新的知识表示,从而实现双语文本情感识别。实验结果显示,与纯中文作为网络输入、纯英文作为网络输入和中英混合文本作为网络输入相比,注意力机制明显优于其他方法;与现有双语情感分析算法相比,该模型有效提升了情感分析的准确率。

The sentiment analysis method of bilingual micro-blog is rare and the accuracy is low,and the same English vocabulary has different effects on the text in different contexts.Therefore,we propose a neural network model for bilingual text sentiment analysis based on attention mechanism.This model used the bi-directional recurrent neural network model to learn the feature representation of text,and introduced the attention mechanism to give different weights to different words in the text,so as to get the new knowledge representation after fusion of features and realize the emotion recognition of bilingual text.The experimental results show that the attention mechanism is superior to other methods compared with pure Chinese as network input,pure English as network input and mixed English text as network input.Compared with the existing bilingual sentiment analysis algorithm,this model improves the accuracy of sentiment analysis.