Explaining Hate Speech Classification with Model-Agnostic Methods (Lecture Notes in Informatics)

“The bidirectional transformer model BERT has been used for prediction because of its state-of-the-art efficiency over other Machine Learning(ML) models. The model-agnostic algorithm LIME generates explanations for the output of a trained classifier and predicts the features that influence the model’s decision. The predictions generated from the model were evaluated manually, and after thorough evaluation, we observed that the model performs efficiently in predicting and explaining its prediction. Lastly, we suggest further directions for the expansion of the provided research work.”

https://recap.uni-trier.de/static/3bb18009cd27ce0999eef0fb60153d61/90.pdf