Exploiting Adaptive Contextual Masking for Aspect-Based Sentiment Analysis

arXiv:2402.13722v1 Announce Type: new
Abstract: Aspect-Based Sentiment Analysis (ABSA) is a fine-grained linguistics problem that entails the extraction of multifaceted aspects, opinions, and sentiments from the given text. Both standalone and compound ABSA tasks have been extensively used in the literature to examine the nuanced information present in online reviews and social media posts. Current ABSA methods often rely on static hyperparameters for attention-masking mechanisms, which can struggle with context adaptation and may overlook the unique relevance of words in …

abstract analysis arxiv cs.cl extraction fine-grained information linguistics literature masking opinions reviews sentiment sentiment analysis social tasks text type