Research on Sentiment Analysis Based on Word2vec and Attention Mechanisml
In allusion to the problems of inaccurate capture of keyword features and incomplete extraction of local sentiment features by traditional sentiment analysis models,resulting in poor classification results,a sentiment analysis model based on TW-BiLSTM-ATT is proposed.Through the improvement of TF-IDF and the combination with Word2vec,the weight feature is inte-grated into the word vector to improve the ability to capture key information.The position feature of the word vector is integrated into the attention mechanism,the model can focus on the words around the target vocabulary,and then extract the emotional features more comprehensively.The comparative experimental results show that the TW-BiLSTM-ATT model has better classification perfor-mance than similar models in processing sentiment analysis tasks.
国家科技期刊平台
NSTL
万方数据
