1. Dataset Collection and Preparation

Combined dataset of toxic and non-toxic comments sourced from publicly available datasets

Convert comments into strings, tokenizing using a tokenizer

Flan-T5-small model, a pre-trained sequence-to-sequence model