This project attempts to build a compact speech recognition system based on RNN-Transducer. It can be used to conduct RNNT-related experiments or to deploy RNNt-based ASR services.
More acoustic models such as Transformer, improved decoders and Linux/Android deployment will be added in the future.