Movie Review Sentiment Analysis Based on CBOW Model and Convolutional Neural Network

Film investors can adjust their film marketing strategies and predict the performance of relevant film markets based on the results obtained from the sentiment analysis of film reviews. The study first obtained more than 1,359 film works and a total of 32,325 movie review data through the crawler, and preprocessed these data. Second, build the CBOW model. After discovering the problems and deficiencies of the model running results, negative sampling and adaptive moment estimation were used to solve the problems of resource exhaustion and low accuracy during the model operation, and the model was optimized. Then, the trained CBOW model was referenced to the convolutional neural network to obtain the fusion model, and after the results and combinations were evaluated through training, the L2 regularization and Adam weight algorithm were used to solve the problem of model overfitting and realize the optimization of the fusion model. It has been verified that the fusion model based on the CBOW model and the convolutional neural network in this study can realize the sentiment analysis of arbitrary input English comments, and then output the predicted sentiment level.