Abstract:The proliferation of hate speech and offensive comments on social media has become increasingly prevalent due to user activities. Such comments can have detrimental effects on individuals' psychological well-being and social behavior. While numerous datasets in the English language exist in this domain, few equivalent resources are available for Persian language. To address this gap, this paper introduces two offensive datasets. The first dataset comprises annotations provided by domain experts, while the second consists of a large collection of unlabeled data obtained through web crawling for unsupervised learning purposes. To ensure the quality of the former dataset, a meticulous three-stage labeling process was conducted, and kappa measures were computed to assess inter-annotator agreement. Furthermore, experiments were performed on the dataset using state-of-the-art language models, both with and without employing masked language modeling techniques, as well as machine learning algorithms, in order to establish the baselines for the dataset using contemporary cutting-edge approaches. The obtained F1-scores for the three-class and two-class versions of the dataset were 76.9% and 89.9% for XLM-RoBERTa, respectively.
Abstract:The large number of exact fitness function evaluations makes evolutionary algorithms to have computational cost. In some real-world problems, reducing number of these evaluations is much more valuable even by increasing computational complexity and spending more time. To fulfill this target, we introduce an effective factor, in spite of applied factor in Adaptive Fuzzy Fitness Granulation with Non-dominated Sorting Genetic Algorithm-II, to filter out worthless individuals more precisely. Our proposed approach is compared with respect to Adaptive Fuzzy Fitness Granulation with Non-dominated Sorting Genetic Algorithm-II, using the Hyper volume and the Inverted Generational Distance performance measures. The proposed method is applied to 1 traditional and 1 state-of-the-art benchmarks with considering 3 different dimensions. From an average performance view, the results indicate that although decreasing the number of fitness evaluations leads to have performance reduction but it is not tangible compared to what we gain.