Natural language processing is a very important and necessary tool to extract valuable information from a large number of comments online. Sentiment analysis and text prediction are both significant parts which have been used in many business areas. It is no doubt that RNN has obtained great progress on natural language processing. In this paper, we used RNN to build models for sentiment analysis and text prediction. Because traditional RNNs cannot handle long-term dependencies, we use LSTM technology to effectively solve this problem. The implement of code is depended on keras. According to the results, the model of sentiment analysis can successfully predict users’ emotion with about 70% accuracy and the model of text prediction provides better context after training many times.