A Hybrid Ensemble Framework for Emotion Detection Using TF-IDF Through Sentiment Analysis: CoRNL
Mrinmoy Kayal, Jayadeep Pati, Himadri Biswas, Abhisek Roy, Rishov Saha, Uddyalok Chakraborty · 2026
Social networks are witnessing an unprecedented surge in demand for text mining applications, particularly in sentiment analysis. With the proliferation of social media, vast volumes of user-generated textual data such as comments and reviews have become readily accessible. Analyzing these texts holds significant value for various business and research applications. Sentiment Analysis (SA), a key technique in Natural Language Processing (NLP), enables the identification of emotions and sentiments embedded in text. This paper introduces the CoRNL model (Combination of Random Forest, Naive Bayes, and Logistic Regression), a hybrid ensemble framework that leverages TF-IDF for feature extraction to predict emotional states from textual data. The proposed methodology involves two primary steps: collecting and labeling the dataset into positive and negative sentiments, and conducting an exhaustive evaluation of four classification models. Our ensemble classifier, CoRNL, demonstrates superior performance compared to individual models Random Forest, Naive Bayes, and Logistic Regression on the Emotion Dataset.