Play with Weighted Finite State Transducers (WFSTs) using the Julia language.

WFSTs provide a powerful framework that assigns a weight (e.g. probability) to conversions of symbol sequences. WFSTs are used in many applications such as speech recognition, natural language processing and machine learning.

This package takes a lot of inspiration from OpenFST.

FiniteStateTransducers is still in an early development stage, see the documentation for currently available features and the issues for the missing ones. An application of WFSTs can be found in the data preparation of the TIDIGITS recipe.