Data Mining Approaches for Sentiment Analysis in Indian Regional Languages: A Review of Challenges, Limitations, and Future Research Directions
TL;DR
This review paper looks at the main data mining methods used for sentiment analysis in Indian regional languages, including machine learning, lexicon-based, rule-based, deep learning, and transformer-based approaches and highlights what they do well and where they struggle.
Abstract
Abstract People across India increasingly use regional languages online to share opinions, reviews, and reactions, but understanding the sentiment behind this text is not easy. Indian regional languages are often used in informal ways, mixed with English, written in different scripts, and supported by only a small number of labeled datasets and language tools. This review paper looks at the main data mining methods used for sentiment analysis in Indian regional languages, including machine learning, lexicon-based, rule-based, deep learning, and transformer-based approaches. It compares how these methods have been used in previous studies and highlights what they do well and where they struggle. The review shows that simple machine learning methods can still be useful when data is limited, while deep learning and transformer models are more effective when better resources are available. Even so, many problems remain, especially code-mixed text, transliteration, sarcasm, negation, and the lack of reliable benchmarks across languages. The paper concludes that future progress depends on building richer datasets, improving language-specific preprocessing, and designing models that are both accurate and practical for real-world use. Keywords: Sentiment Analysis; Data mining; Indian regional languages; low-resource NLP; code-mixing