← Back⌂ All journals
Review of Artificial Intelligence in Educatione-ISSN 2965-4688v. 8 (2025)Open Access1 min read

A Comprehensive Review of Current Methods, Progress, and Challenges in Sentiment Analysis

  1. INMaharashtra Institute of Technology (MIT)
Read as flipbookRead PDFAll articles in this issueCite this article

Abstract

Purpose

This paper reviews the current state, progress, and challenges of Sentiment Analysis (SA), a computational approach to identifying and interpreting human opinions and emotions from text. It aims to evaluate the evolution of methods, highlight persistent limitations, and explore emerging directions toward developing generalizable, explainable, and ethically responsible sentiment analysis systems.

Design / Methodology / Approach

The review integrates evidence from two key sources: (1) a tertiary study (Lighart et al., 2021) summarizing outcomes from 14 systematic literature reviews and mapping studies, and (2) a domain-specific review (Sweta, 2024) on sentiment analysis in education. Through thematic analysis, the study identifies methodological trends, core challenges, and innovations shaping the discipline, linking large-scale findings with domain-level applications.

Findings

The analysis reveals a significant methodological transition—from lexicon-based and traditional machine learning models (SVM, Naive Bayes) to deep learning and transformer-based architectures (CNN, LSTM, BERT, GPT). Despite progress, key issues persist: domain and language dependency, lack of contextual sensitivity (sarcasm, irony), data imbalance, model opacity, and privacy concerns. New directions, such as cross-domain and multilingual modeling, explainable AI (XAI), multimodal analysis, and real-time SA, demonstrate promising solutions.

Research Limitations / Implications

The review highlights the need for cross-lingual datasets, explainable architectures, and interdisciplinary collaborations to enhance generalizability, fairness, and ethical compliance.

Originality / Value

By combining high-level synthesis with domain-specific insights, this work provides a comprehensive and forward-looking framework for building robust, transparent, and socially responsible sentiment analysis models.

Keywords
Sentiment analysisOpinion miningDeep learningExplainable AIMultilingual NLP

Cite this article

BibTeXRIS

Zotero and Mendeley: open the RIS file, or use the Zotero Connector on this page (it reads the article metadata).

Share

Related articles