Artificial Intelligence Technology-Based Semantic Sentiment Analysis on Network Public Opinion Texts

Abstract

Considering that the current social network text analysis works poorly in accurate and effective sentiment prediction and management, a deep learning (D-L)-based text sentiment analysis method is proposed for the big data environment. First, the autoregressive language model mode XLNet is used to capture bidirectional text information and a sentiment analysis model XLNet-Multi-Attention-BiGRU. Then, considering the context information of social network texts, the defect of traditional GRU units only reading texts in order is overcome by introducing a BiGRU model to extract features in both directions. Finally, a multi-headed attention layer is added between the BiGRU and CRF layers to better capture the key information in the sentence by integrating multiple single-head attention. The results show that the precision, recall, and F1 value of the method proposed in this paper are the largest, with the highest reaching 92.64%, 92.32%, and 91.25%, respectively, which are 12.40%, 10.17%, and 9.63% higher than the maximum values of the other three methods, respectively.