文献类型：期刊文献

英文题名：A Sentiment Analysis Model Based on Text Generation-OpnionSpanT5

作者：Liu, Peiyu[1]; Guo, Weibin[1]

机构：[1] School of Information Science and Engineering, East China University of Science and Technology, Shanghai, China

年份：2024

起止页码：1536

外文期刊名：2024 IEEE 2nd International Conference on Control, Electronics and Computer Technology, ICCECT 2024

收录：EI(收录号：20242516268355)

语种：英文

外文关键词：Deep learning

摘要：ABSA is a type of fine-grained sentiment analysis, aiming to analyze the various sentiment types of a given sentence in more detail. Currently, more and more solutions adopt Seq2Seq-based architecture to perform sentiment analysis by outputting a sentence containing the required sentiment tuples. However, there is usually a problem, that is, when capturing opinion items, there may be a problem of missing extraction when facing opinion items with span. This paper proposes a model based on the paraphrase generation method, that is, before generating the quadruple, to extract opinion items separately. A BiLSTM-CRF layer is added in front of the PARA model to extract opinion items separately, and then the opinion items and the original sentences are input into the subsequent T5 model to generate emotional interpretations. We call it OpinionSpanT5, which is effective Fixed an error that occurred when generating opinion items. ? 2024 IEEE.