Research on Text Sentiment Analysis of Movie Reviews Based on BERT Model
Deqing Zhang, Cuoling Zhang · 2023
With the rapid development of machine learning technology, more and more industries pay attention to the research and application of data science. In recent years, many industries such as e-commerce, service, movie and television, news and so on have begun to attach great importance to and deeply analyze the sentimental tendency of user comments in the network. Sentiment analysis is an important branch of NLP. This paper takes the movie reviews dataset published by Stanford as the research object, and uses the bert-base-uncased pre-training model in BERT to study the text sentiment classification of the dataset. In the fine tune phase of the model, Adam and CrossEntropyloss are selected as the optimizer and loss function, and at the same time, the influence of parameters learning_rate and batch_size on the accuracy of text classification is studied. The experimental results show that learning_rate and batch_size has a certain impact on classification accuracy. When learning_rate=2e-5, the classification accuracy of the model on this dataset can reach up to 87%, achieving a relatively ideal classification effect.