Zhang, Wenxuan, et al. "Aspect-based sentiment analysis in question answering forums." Findings of the Association for Computational Linguistics: EMNLP 2021. 2021.

Aspect-based sentiment analysis (ABSA) traditionally focuses on extracting aspects and predicting sentiments from individual sentences, such as in customer reviews. This paper explores ABSA in the context of question-answer (QA) forums, where opinions are distributed across question and answer pairs. The proposed model introduces a cross-sentence aspect-opinion interaction mechanism to better align aspects mentioned in questions with the corresponding opinions in answers. Evaluations on three real-world datasets demonstrate that this model significantly outperforms existing state-of-the-art methods.