Feature-Based Linguistic Text Sentiment Analysis Using Stacked Meta-Ensemble Learning

Sentiment analysis, or opinion mining, is a crucial aspect of Natural Language Processing (NLP) applications. Transformer-based models are commonly used, but their accuracy is limited in languages with limited linguistic resources. Recent studies have combined ensemble learning approaches with deep learning techniques to improve sentiment analysis. This research article focuses on sentiment analysis in the Indian linguistic context. This paperproposes an ensemble architecture for linguistic feature- based sentiment analysis, including a meta-learner, which improves accuracy and recall rates by combining Bert modelsand integrating features into meta-data-based models. The proposed model achieves good accuracy-recall ratings andoutperforms baseline deep learning models, highlighting its effectiveness.