Xinyuan Li d305c7cceb Implement recipe for Fluent Speech Commands dataset
Signed-off-by: Xinyuan Li <xli257@c13.clsp.jhu.edu>
2024-01-19 13:37:00 -05:00

9 lines
573 B
Markdown
Executable File

## Fluent Speech Commands recipe
This is a recipe for the Fluent Speech Commands dataset, a speech dataset which transcribes short utterances (such as "turn the lights on in the kitchen") into action frames (such as {"action": "activate", "object": "lights", "location": "kitchen"}). The training set contains 23,132 utterances, whereas the test set contains 3793 utterances.
Dataset Paper link: <https://paperswithcode.com/dataset/fluent-speech-commands>
cd icefall/egs/fluent_speech_commands/
Training: python transducer/train.py
Decoding: python transducer/decode.py