Sentiment Analysis of Movie Reviews Based on Improved Word2vec and Ensemble Learning

Abstract

Taking the field of movie review as an example, this paper proposes a sentiment analysis method based on improved word2vec and ensemble learning. The basic design idea is: firstly build the corresponding corpus through new word discovery and use the TF-IDF algorithm to exponentially weight the word2vec word vector, which is used to integrate the semantic relationship between words and the importance of vocabulary information into the model; secondly, to avoid the cumbersome problems of data labeling, the existing algorithms of automatic labeling reviews are improved to increase the adoption rate of data; finally, Stacking algorithm is used to train and classify the emotional data. The proposed model can simplify the domain text representation and improve the classification performance of the model. The experimental results show that compared with existing methods, the accuracy, precision and recall rate of the algorithm proposed in this paper have been improved on film review data.