Research on Improved ABSA: Synergistic Effect of Sentence Splitting and Data Enhancement

The task of Sentiment quadruple extraction in Aspect-Based Sentiment Analysis (ABSA) faces two core chal-lenges: the sparsity of labeled data and the complex semantic parsing of compound sentences. To address the above challenges, this study proposes a Sentence Splitting Based on Aspect-Inverse Text Data Augmentation (SSB-ITDA) model. This paper aims to solve the problem of data scarcity in the process of sentence division of compound sentences. In this study, SSB was used to parse aspect terms in sentiment quadruples and split complex compound sentences into simple clauses with clear semantics. At the same time, ITDA is used, and the generation strategy of quadruple-to-text is adopted to expand high-quality labeled data, alleviate the problem of data scarcity, and make sentence segmentation more accurate. Experimental results show that the proposed method is significantly better than the existing methods on multiple datasets, which effectively verifies the collaborative optimization effect of data augmentation and sentence segmentation module.