Point dataset reference to renamed billingsmoore/nict-tib1 repo
Browse files
README.md
CHANGED
|
@@ -9,7 +9,7 @@ tags:
|
|
| 9 |
- whisper
|
| 10 |
base_model: openai/whisper-small
|
| 11 |
datasets:
|
| 12 |
-
- billingsmoore/
|
| 13 |
metrics:
|
| 14 |
- cer
|
| 15 |
- wer
|
|
@@ -20,7 +20,7 @@ model-index:
|
|
| 20 |
type: automatic-speech-recognition
|
| 21 |
name: Automatic Speech Recognition
|
| 22 |
dataset:
|
| 23 |
-
type: billingsmoore/
|
| 24 |
name: NICT-Tib1 (Lhasa Tibetan)
|
| 25 |
split: test
|
| 26 |
metrics:
|
|
@@ -46,7 +46,7 @@ Whisper is an encoder–decoder transformer, originally pre-trained by OpenAI on
|
|
| 46 |
|
| 47 |
## Training data
|
| 48 |
|
| 49 |
-
Fine-tuned on [NICT-Tib1](https://huggingface.co/datasets/billingsmoore/
|
| 50 |
|
| 51 |
## Training procedure
|
| 52 |
|
|
|
|
| 9 |
- whisper
|
| 10 |
base_model: openai/whisper-small
|
| 11 |
datasets:
|
| 12 |
+
- billingsmoore/nict-tib1
|
| 13 |
metrics:
|
| 14 |
- cer
|
| 15 |
- wer
|
|
|
|
| 20 |
type: automatic-speech-recognition
|
| 21 |
name: Automatic Speech Recognition
|
| 22 |
dataset:
|
| 23 |
+
type: billingsmoore/nict-tib1
|
| 24 |
name: NICT-Tib1 (Lhasa Tibetan)
|
| 25 |
split: test
|
| 26 |
metrics:
|
|
|
|
| 46 |
|
| 47 |
## Training data
|
| 48 |
|
| 49 |
+
Fine-tuned on [NICT-Tib1](https://huggingface.co/datasets/billingsmoore/nict-tib1), a corpus of transcribed Lhasa Tibetan speech from 20 speakers (Soky, Gong & Li, 2022). The data was split by speaker (85/15, not by utterance) so that no speaker appears in both splits: 15,099 training utterances from 17 speakers, 1,547 test utterances from the remaining 3 speakers.
|
| 50 |
|
| 51 |
## Training procedure
|
| 52 |
|