Ensemble Approaches for Offensive Language Detection in Kannada

In the realm of natural language processing, sentiment analysis remains a focal point due to its pivotal role in discerning and analyzing public sentiments. While past research has been dedicated to enhancing the performance of sentiment analysis models, this study delves into the potential benefits of ensemble techniques, with a special emphasis on Genetic Algorithm (GA) optimization, The main thrust of this research was to investigate whether merging multiple models and subsequently optimizing their weights via GA could elevate performance. Evaluations were thoughtfully conducted, focusing on the weighted precision, recall, and F1-scores of individual models. Additionally, the study compared the ensemble methods, both with and without GA optimization, The ensemble method, even without the GA optimization, secured an impressive F1-score of 0.73. Upon integrating the GA optimization, there was a gentle improvement in the ensemble’s performance, achieving an F1-score of 0.75, In summing up, this study manifests a palpable enhancement in sentiment analysis outcomes when shifting from standalone models to a GA-optimized ensemble mechanism. The results, epitomized by an F1-score peaking at 0.75, underline the intrinsic merits of harnessing GA for weight optimization in ensemble constructs.