EUVIMEDEUVIMED entdecken
Crossref

A Comprehensive Review of Multilingual Sentiment Analysis for Indian Languages from Traditional Machine Learning to Large Language Models

Meenal M. Shingare, Dr. B. S. Sonawane Dr. B. S. Sonawane

IOSR Journal of Computer Engineering · 2026-03 · 28 (3) · 22-48

DOI: 10.9790/0661-2803032248 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IOSR Journal of Computer Engineering, 2026-03. Verfasst von Meenal M. Shingare und 1 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Sentiment analysis is an important research area in Natural Language Processing (NLP) which is the automated identification and interpretation of opinions, emotions and attitudes expressed in text data. The growth of user generated content on social media, review systems, discussion forums and digital communication channels has created the demand for accurate and scalable sentiment analysis systems. There has been a lot of work on sentiment analysis in English language, but the development of effective models for multilingual and low-resource languages still remains an intimidating research problem. The present work provides a comprehensive review of the literature on sentiment analysis from 2020 to 2026, focusing on multilingual sentiment analysis, sentiment analysis for Indian languages, transformer-based models, and recent large language models. A systematic literature review methodology based on the PRISMA framework was applied to identify, screen and analyze relevant studies from major scientific databases including Scopus, IEEE Xplore, Web of Science, ACM Digital Library, SpringerLink and ScienceDirect. The review categorizes the existing works into five paradigms, namely, lexicon-based, machine learning, deep learning, transformer-based and multilingual sentiment analysis. Special focus is placed on the Hindi and Marathi sentiment analysis resources including benchmark datasets such as IIT Patna Movie Reviews, L3CubeMahaSent, MahaSent-MD and code-mixed social media corpora. The review also includes the evolution of the text representation techniques from Bag-of-Words, TF-IDF to contextual embeddings generated by BERT, IndicBERT, MuRIL and large language models. Transformer-based architectures consistently outperform traditional machine learning and deep learning approaches on various sentiment analysis benchmarks, as the analysis shows. But issues such as code-mixing, transliteration, low-resource datasets, explainability, computational complexity and bias are still largely unresolved. The review discusses emerging directions such as multilingual transformers, explainable artificial intelligence, multimodal sentiment analysis, retrieval-augmented generation, federated learning and large language models and identifies significant research gaps. The results can serve as a comprehensive reference for researchers and practitioners involved in the development of next generation multilingual sentiment analysis systems, supporting various language and realworld application environments.

Zitieren

APA

Shingare, M. M., & Sonawane, D. B. S. S. D. B. S. (2026). A Comprehensive Review of Multilingual Sentiment Analysis for Indian Languages from Traditional Machine Learning to Large Language Models. IOSR Journal of Computer Engineering, 28(3), 22-48. https://doi.org/10.9790/0661-2803032248

Vancouver

Shingare MM, Sonawane DBSSDBS. A Comprehensive Review of Multilingual Sentiment Analysis for Indian Languages from Traditional Machine Learning to Large Language Models. IOSR Journal of Computer Engineering. 2026;28(3):22-48. doi:10.9790/0661-2803032248

BibTeX
@article{Shingare2026,
  author = {Shingare, Meenal M. and Sonawane, Dr. B. S. Sonawane Dr. B. S.},
  title = {A Comprehensive Review of Multilingual Sentiment Analysis for Indian Languages from Traditional Machine Learning to Large Language Models},
  journal = {IOSR Journal of Computer Engineering},
  year = {2026},
  volume = {28},
  number = {3},
  pages = {22--48},
  doi = {10.9790/0661-2803032248}
}
Kostenlos starten →