Xiangyang Feng

BERT-based short text sentiment classification model integrating dynamic pruning and multiscale attention

To address the problems of large parameter scale, low inference efficiency, and insufficient capture of local sentiment cues in the original BERT model for short-text sentiment classification, this paper proposes an improved BERT-based sentiment classification model integrating dynamic pruning and multi-scale attention...