To rapidly capture public opinion trends on online social platforms during specific earthquake events, sentiment analysis plays a crucial role in public opinion monitoring. To enhance the efficiency of sentiment analysis for earthquake-related online discourse, this study proposes a sentiment analysis model based on Long Short-Term Memory (LSTM) networks. First, earthquake-related public opinion data are collected using web-crawling programs and subjected to standard preprocessing. The text is then converted into word vectors using Word2Vec, after which an LSTM-based sentiment analysis model tailored to earthquake-related public opinion is constructed to classify posts into positive, negative, and neutral categories. A series of experiments verifies the effectiveness of the proposed model.