Sentiment Analysis of Movie Reviews Based on Sentiment Dictionary and Deep Learning Models

As the Internet era has progressed, platforms such as Douban Movies have spawned a great number of evaluations with personal biases. However, these assessments lack a set length, and the text's expression is varied, not constrained to grammar-related constraints. The expressive style is less formal. As a result, mining and assessing these opinions has substantial economic worth. This experiment utilized a novel sentiment lexicon to adapt informal vocabulary in movie reviews. To improve the accuracy of sentiment analysis in movie reviews, it was integrated with the Albert-BiLSTM-Attention model. The results of six rounds of comparative experiments show that the method suggested in this paper has improved average precision, average recall, and average F1 score in the sentiment classification of this dataset. The suggested model can be used to achieve precise sentiment analysis for film reviews, offering pertinent support and advice for the production team's upcoming films.