Research and Analysis in Fine-grained Sentiment of Film Reviews Based on Deep Learning

Films are one of the main entertainment ways for people. By exploring, analyzing and summarizing the reviews that people published on the internet, audiences can make better viewing choices, while investors can get a more convenient way to understand the audience's feedback. The text proposes a method of deep learning to perform fine-grained sentiment analysis on the film reviews, and restores the user's real emotion as much as possible. The method first preprocesses the data, and converts the words into vectors using the Word2Vec model, then inputs the word sequence in the Long Short-Term Memory network (LSTM) to learn the semantic dependence. Finally, the logical regression classifier is used to classify. Compared with Naive Bayes (NB), Support Vector Machine (SVM), K-Nearest Neighbor (KNN), Back Propagation Neural Network (BPNN), and Convolutional Neural Network (CNN) models, the method achieves 83.6% accuracy in the binary classification, and 76.1% and 51.2% accuracy in the positive and negative fine-grained emotional classification, which has achieved the best results proved by experiments.