Juicer is a Weighted Finite State Transducer (WFST) based decoder for Automatic Speech Recognition (ASR).
It was developed at Idiap under the AMI and AMIDA projects in collaboration with the universities of Sheffield and Edinburgh.
Being WFST based, Juicer is capable of large vocabulary decoding as well as smaller command and control tasks. There is a paper about Juicer on the Idiap publications site:
Juicer: A Weighted Finite-State Transducer speech decoder
Recently, the code base has been moved to GitHub.