Tokenization is a essential method in machine learning, acting as the first stage for preparing text data. Essentially, it involves splitting up a larger block of information – like a sentence – into smaller units called “pieces". These tokens can be terms, symbols, or even sub-lexical segments.