Humor Detection Using Machine Learning Approach

Humor, as a verbal expression that frequently conveys meanings contradictory to the literal words, presents a hard barrier for machines to decipher. Its core is found not just in words but also in the subtleties of intonation, irony, and contextual nuances, making computational comprehension a difficult endeavor. This study intends to dig into the field of humor detection using machine learning approaches. We investigate various machine learning models designed for Humor Detection in this paper. Logistic Regression, Decision Tree, Random Forest, Support Vector Machine, Naive Bayes, XGBoost, ADA Boost, and Gradient Boosting are among the models included. Among these models, the ensemble technique of Logistic Regression, Mini Batch K-Means and Gradient Boosting stands out as a no-table performer. This innovative approach outperforms standard algorithms with a surprising 93 % accuracy in testing and an astounding 94 % accuracy in training. This achievement throws light on the possibility of more effective and efficient humor detection techniques, indicating a substantial advancement in computational humor analysis.