Social Media Hate Speech Detection Using Explainable Artificial Intelligence (XAI)

Explainable artificial intelligence (XAI) characteristics have flexible and multifaceted potential in hate speech detection by deep learning models. Interpreting explaining decisions made complex (AI) models to understand the decision-making process of these model were aims this research. As a part research study, two datasets taken demonstrate using XAI. Data preprocessing was performed clean data any inconsistencies, text tweets, tokenize lemmatize text, etc. Categorical variables also simplified order generate dataset for training purposes. Exploratory analysis on uncover various patterns insights. Various pre-existing applied Google Jigsaw such as decision trees, k-nearest neighbors, multinomial naïve Bayes, random forest, logistic regression, long short-term memory (LSTM), among which LSTM achieved an accuracy 97.6%. methods LIME (local interpretable model—agnostic explanations) HateXplain dataset. Variants BERT (bidirectional encoder representations from transformers) + ANN (artificial neural network) with 93.55% MLP (multilayer perceptron) 93.67% created achieve good performance terms explainability ERASER (evaluating rationales simple English reasoning) benchmark.