Yanzhi Xu, Yueying Hua, Shichen Li, Zhongqing Wang. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Aspect-based sentiment analysis (ABSA) aims to extract the aspect terms from review text, and to predict the polarity towards the aspect term. Although neural models have achieved competitive results, there are still many challenges in this task. Firstly, there is irrelevant and noise information in the review text, and the offsets of aspect term boundary are hard to decide. In addition, sentiment is usually either expressed implicitly or shifted due to the occurrence of negation and rhetorical words. To tackle the above limitations, we propose a scope detection model to distinguish whether the words from the review text are relevant with the aspect term, and to filter irrelevant and noise information. In addition, we investigate a biaffine-based model to constrain the scope detection process of aspect term extraction. We further generate a simplified clause based on the scope of aspect term, and predict the polarity based on the simplified clause. Empirical studies show the effectiveness of our proposed model over several strong baselines. These also justify the importance of scope detection for aspect-based sentiment analysis.

Stance detection aims to assign a stance label (in favor or against) to a post towards a specific target. In the literature, there are many studies focusing on this topic, and most of them treat stance detection as a supervised learning task. Therefore, a new classifier needs to be built from scratch on a well-prepared set of ground-truth data whenever predictions are needed for an unseen target. However, it is difficult to annotate the stance of a post, since a stance is a subjective attitude towards a target. Hence, it is necessary to learn the information from unlabeled data or other target data to help stance detection with a certain target. In this study, we propose a multi-target stance detection framework to integrate multi-target data together for stance detection. Since topic and sentiment are two important factors to identify the stance of a post in multi-target data, we propose an adversarial attention network to integrate multi-target data by detecting and connecting topic and sentiment information. In particular, the adversarial network is utilized to determine the topic and the sentiment of each post to collect some target-invariant information for stance detection. In addition, the attention mechanism is utilized to connect posts with a similar topic or sentiment to acquire some key information for stance detection. The experimental results not only demonstrate the effectiveness of the proposed model, but also indicate the importance of the topic and the sentiment information for stance detection using multi-target data.