|
--- |
|
language: |
|
- en |
|
tags: |
|
- Voice |
|
datasets: |
|
- JBJoyce/DENTAL_CLICK |
|
metrics: |
|
- accuracy |
|
--- |
|
|
|
### Model Description |
|
Model utilizes Wav2vec2 architecture trained on the Superb dataset for keyword spotting task and was fine |
|
tuned to identify dental dental click utterance (https://en.wikipedia.org/wiki/Dental_click) in speech. |
|
Model was trained for 10 epochs on a limited quantity of speech (~1.5 hours) and with only one speaker. |
|
Thus the model should not be assumed to hold generalizability to other speakers or languages without further |
|
training data or rigorous testing. |
|
|
|
Model was evaluated for accuracy on a hold out test set of 20% of the available data and scored 97%. |
|
|
|
## Uses |
|
Model can be used via transformers library or via Hugging Face Hosted inference API to the right. I would |
|
caution against the use of the 'Record from browser' option as model may erronously identify user's mouse |
|
click as a speech utterance. Audio files for upload should be 1 sec in length, with 'WAV' format and 16 bit |
|
signed integer PCM encoding. |
