A survey on aspect base sentiment analysis methods and challenges

Abstract

Sentiment analysis has spread to virtually every industry, including service, finance, e-commerce, consumer goods, telecommunications, health care, political campaigns, social events, and elections. Aspect-based sentiment analysis (ABSA) enables the automatic extraction of sentiment information from textual information or phrases that are deeply embedded. To address ABSA in various situations, numerous tasks for assessing various sentiment components and associated relationships. Our proposed ABSA workflow divides into broad categories: ABSA task as a single and compound, deep learning approach for ABSA. Specifically, we present a novel taxonomy for ABSA that organizes existing studies based on the axes of sentimental components of relevance, with a focus on modern improvements in complex ABSA tasks. In contrast to earlier ABSA studies that focused on a single sentiment element, numerous ABSA tasks involving numerous components have been examined in recent years in order to capture more comprehensive aspect-level sentiment polarity. Nonetheless, a comprehensive examination of the various ABSA objectives and their corresponding results is still lacking, a gap that this survey aims to address. In addition, trends in ABSA research are noted, and how the ABSA area can be progressed in the future prospective.