Sentiment analysis of twitter user comments based on long short-term memory networks

With the proliferation of social media, Twitter, as an important information dissemination platform, contains rich emotional information in user comments. In recent years, deep learning, especially Long Short-Term Memory Networks (LSTM), has achieved significant success in sentiment analysis. However, challenges remain in processing short texts and the non-standard, character-limited nature of Twitter comments. This study aims to analyze the sentiment of Twitter user comments using LSTM. So a neural network model was constructed, including an embedding layer, dropout layer, bidirectional LSTM layer, and fully connected layer, and trained using the Adam optimizer. The study found that for comments with simple words and case-sensitive vocabulary, the sentiment analysis model achieved an accuracy of about 98%. However, for comments containing emojis and complex symbols, the accuracy dropped to 91%. Overall, the study results indicate that the model can accurately identify the emotional tendencies in comments and adapt well to different types of comments, achieving an overall accuracy of 88% on the test set. Additionally, the model maintains a high accuracy for comments with emojis and special symbols, and it is slightly higher for simple short sentences and case-sensitive comments. By adjusting parameters and optimizing algorithms, the model's performance was successfully improved. This study provides important empirical evidence and guidance for sentiment analysis of Twitter user comments based on LSTM, which has positive implications for improving and optimizing related models.