Given an utterance the spellcheck corrects each token based on a frequency dictionary. It is language agnostic that means it works with any language as soon as a frequency dictionary is provided. The approach is based on the Peter Norvig's blog post on the spell check (see: https://norvig.com/spell-correct.html)