This repository implements speech recognition using Contextnet-CTC.
Contextnet is a speech recognition model using SEnet.
Zeroth dataset was used for the experiment. Zeroth data is 51.6 hours transcribed Korean audio for training data and 1.2 hours transcribed Korean audio for testing data.
python train.py --conf config/contextnet_ctc.yaml
python evaluate.py --conf config/contextnet_ctc.yaml
| Data | Model | CER | Decoding |
|---|---|---|---|
| Zeroth test data | Contextnet-CTC | 9.58% | beam search |