22
- Volume 3, Issue 2 2025
By Afsheen Maroof , Shaukat Wasi
10.20547/aibd.253201
Keywords: Keywords: systematic literature review, ABSA task, ABSA datasets, ABSA trends, opinion term extraction, sentiment classification, aspect categorization, sentiment Analysis
Abstract: With the expansion of online platforms, new rooms are opening to extract full insights from huge textual data. Aspect-Based Sentiment Analysis (ABSA) is rising as an essential Artificial Intelligence (AI) discipline that employs Natural Language Processing (NLP) to associate sentiments with specific attributes of entities. In this systematic review, we focused on ABSA studies published between 2020 and 2024, encompassing 66 primary research articles focusing on advancements in AI methodologies, including transformer-based models and generative AI techniques. The novel ABSA taxonomy developed in this study defines methodological categories and analyzes the use of commercial and domain-specific datasets. This research emphasizes the practical need for domain-specific datasets in ABSA. for domain domain-specific dataset for ABSA. It also reveals significant trends, including the growing acceptance of deep learning (DL), machine learning (ML), and LLM-based/ transformer models. This paper also highlights limitations in current research and provides actionable insights for advancing ABSA methodologies, such as encouraging the use of more robust generative models for ABSA tasks and outlining strategies for developing new domain-specific datasets. These contributions can be used as a reliable resource for academicians and researchers for reference in the domain of sentiment analysis.
Submission Date: 1 Jul, 2025 Reviews Completed: 3 Oct, 2025Acceptance Date: 24 Nov, 2025 Publication Date: 31 Dec, 2025