Open AccessA Practical Application for Text-Based Sentiment Analysis Based on Bayes-LSTM Model
Author(s) -
Jiawen Li,
Huaping Zhu
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1631/1/012035
Subject(s) - computer science , naive bayes classifier , artificial intelligence , preprocessor , microblogging , sentiment analysis , classifier (uml) , data pre processing , social media , word (group theory) , machine learning , web crawler , natural language processing , feature selection , data mining , support vector machine , world wide web , linguistics , philosophy
Text-based sentiment analysis algorithms have now become one of the active research areas in emotional analysis which has gained much attention nowadays. Text emotion classification can be widely used in social public opinion analysis, product use feedback, harmful information filtering, etc. In this paper, we first developed a robotic crawler to gather data about comment on Huawei cellphone from Sina weibo microblog sites (Chinese twitter). Then we generate the data text to be trained according to the input requirements of the Keras module, and perform formal training and learning on the model after data preprocessing. Subsequently, the classifier was constructed based on the Bayes-LSTM model in which TF-IDF model was used for feature selection. The LSTM model can be characterized by the ability to self-evaluate the usefulness of the information obtained, which makes up for the shortcoming of naive Bayes formula that only applies to two independent events. We finally have a practical application that generates a word cloud from text, showing frequently used words in larger font sizes, effectiveness of the algorithm was also verified by experiment.