The wealth of unstructured text on the online web portal has made opinion mining the most thrust area for researchers, academicians, and businesses to extract information for gathering, analyzing, and aggregating human emotions. The extraction of public sentiment from the text at an aspect level has contributed exceptionally to various businesses in the marketplace. In recent times, deep learning-based techniques have learned high-level linguistic features without high-level feature engineering. Therefore, this paper focuses on a rigorous survey on two primary subtasks, aspect extraction and aspect category detection of aspect-based sentiment analysis (ABSA) methods based on deep learning. The significant advancement in the ABSA sector is demonstrated by a thorough evaluation of state-of-the-art and latest aspect extraction methodologies.

在线门户网站上大量的非结构化文本使得观点挖掘成为研究人员、学者和企业提取信息以收集、分析和聚合人类情感的最热门领域。从文本中提取公众情绪在一定程度上对市场上的各种业务做出了巨大贡献。近年来，基于深度学习的技术无需高级特征工程即可学习高级语言特征。因此，本文重点对基于深度学习的基于方面的情感分析（ABSA）方法的两个主要子任务：方面提取和方面类别检测进行了严格的调查。对最先进和最新方面提取方法的全面评估证明了 ABSA 领域的重大进步。