Tokenization is a essential method in machine learning, acting as the initial stage for preparing text data. Essentially, it involves dividing a larger block of text – like a phrase – into smaller units called “segments". These tokens can be strings, punctuation marks, or even sub-lexical segment