Sarcasm Detection in News Headline using Stack Ensemble Technique
Mohammed Ali Kawo, Garba Muhammad, Danlami Gabi, Musa Sule Argungu · International Journal of Computer Science and Mobile Computing · 2025
The challenges and benefits of sarcasm detection in opinion mining has led to curiosities in researches; to distinguish sarcasm in a sentence has become strenuous task for humans and it causes a lot of conflicts when analysing sentiments, text analysis and in computational linguistics. The individual base classifiers tend to predict in a biased form with a decline in accuracy scores. Sarcasm obstructs a lot of effective models in machine learning techniques. This research stresses in addressing the challenges of sarcasm and to effectively generate a Stack ensemble technique to detect sarcasm in news headline dataset. Naive Bayes (NB), Support Vector Machine (SVM), K-Nearest Neighbour (KNN), Decision Tree (DT) and a Logistic Regression (LR) Meta-Stack ensemble classifier were classified with the following metrices such as accuracy, precision, recall and F1-scores on a news headline dataset from kaggle.com to ascertain the best performance classifier in dealing with sarcasm detection. The stack ensemble model outperforms all other classifiers by achieving a significant accuracy score of 87%, precision for no-sarcasm is 85%, while for sarcasm is 81%, recall for no-sarcasm is 86%, while for sarcasm is 79% and the F1-score for no-sarcasm is 85% and for sarcasm is 80%. These revealed results have depicted how effective stack ensemble model performed in detecting sarcasm in news headline dataset. However, it is essential to exploit other advanced techniques such as deep learning in respect of sarcasm, so as to improve in accuracy scores.