Research And Design Of Sentiment Analysis Of Weibo Comment Text Based On Deep Learning

In recent years,social platforms led by Weibo have achieved rapid development due to the development of the mobile Internet.In just a decade of development,its users have grown from more than one million users where the platform was established to nearly now.The data generated by one billion users can no longer be ignored.This article mainly focuses on the text sentiment analysis of the comments under the hot blog posts on Weibo.Through the sentiment analysis of these very subjective short texts,we can understand the views and attitudes of the general public on the Internet hot public opinion events.Reasonable use of these analysis results can effectively supervise and manage public opinion,which is conducive to relevant departments to formulate relevant policies more flexibly and better serve the people.This paper analyzes the three typical development status of text sentiment analysis and its advantages and disadvantages,and finally adopts the deep learning method to carry out the experimental design of this paper.Combined with the pretraining model,Albert can generate different word vectors in different contexts based on the same word.Bi LSTM,a bidirectional long and short-term memory neural network,can read text data from front to back and from back to front at the same time,with enhanced semantic pairs The characteristics of the degree of context dependence and the attention mechanism Attention can ignore non-important information in a large amount of information,selectively refine a small amount of important information and give it a higher weight,and have the characteristics of focusing on important information and ignoring non-important information.The Albert-Bi LSTM-Attention model is constructed,and the optimal parameters suitable for this article are obtained during the training process of the model.After data acquisition and text preprocessing,the preprocessed data is input to the four groups of models of Albert-Bi LSTMAttention,Albert-LSTM-Attention,Albert-Bi LSTM and Skip-Gram-Bi LSTMAttention models for comparative experimental verification.According to the experimental results on two different data sets,the model constructed in this paper has a good effect from the three evaluation index results of accuracy rate,recall rate and F1.