Updated README
Browse files
README.md
CHANGED
@@ -9,16 +9,27 @@ pipeline_tag: automatic-speech-recognition
|
|
9 |
|
10 |
# General-purpose Latvian ASR model
|
11 |
|
12 |
-
This is a fine-tuned [whisper-large-v3](https://huggingface.co/openai/whisper-large-v3) model for Latvian, trained by [AiLab.lv](https://ailab.lv) using two general-purpose speech datasets: the Latvian part of [Common Voice 17.0](https://commonvoice.mozilla.org/lv/datasets)
|
13 |
|
14 |
We also provide a quantized version of the model in the GGML format.
|
15 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
16 |
## Evaluation
|
17 |
|
18 |
-
The model is evaluated on both a Latvian CV 17.0 test set (the V1 split) and a [LATE-Media test set](http://hdl.handle.net/20.500.12574/99).
|
19 |
|
20 |
-
|
|
21 |
-
|
22 |
| Latvian CV 17.0 V1 - formatted | 5.0 | 1.6 |
|
23 |
| Latvian CV 17.0 V1 - normalized | 3.4 | 1.0 |
|
24 |
| LATE-Media 1.0 - formatted | 20.8 | 8.2 |
|
|
|
9 |
|
10 |
# General-purpose Latvian ASR model
|
11 |
|
12 |
+
This is a fine-tuned [whisper-large-v3](https://huggingface.co/openai/whisper-large-v3) model for Latvian, trained by [AiLab.lv](https://ailab.lv) using two general-purpose speech datasets: the Latvian part of [Common Voice 17.0](https://commonvoice.mozilla.org/lv/datasets), and a Latvian broadcast dataset [LATE-Media](https://korpuss.lv/id/LATE-mediji).
|
13 |
|
14 |
We also provide a quantized version of the model in the GGML format.
|
15 |
|
16 |
+
## Training
|
17 |
+
|
18 |
+
Fine-tuning was done using the Hugging Face Transformers [library](https://github.com/huggingface/transformers/tree/main/examples/pytorch/speech-recognition#sequence-to-sequence).
|
19 |
+
|
20 |
+
Training data:
|
21 |
+
| Dataset | Hours |
|
22 |
+
|:---|---:|
|
23 |
+
| Latvian CV 17.0 train set (the [V1 split](https://analyzer.cv-toolbox.web.tr/examine/lv/17.0)) | 167 |
|
24 |
+
| LATE-Media train set | 42 |
|
25 |
+
| Total | 209 |
|
26 |
+
|
27 |
## Evaluation
|
28 |
|
29 |
+
The model is evaluated on both a Latvian CV 17.0 test set (the [V1 split](https://analyzer.cv-toolbox.web.tr/examine/lv/17.0)) and a [LATE-Media test set](http://hdl.handle.net/20.500.12574/99).
|
30 |
|
31 |
+
| Dataset | WER | CER |
|
32 |
+
|:---|---:|---:|
|
33 |
| Latvian CV 17.0 V1 - formatted | 5.0 | 1.6 |
|
34 |
| Latvian CV 17.0 V1 - normalized | 3.4 | 1.0 |
|
35 |
| LATE-Media 1.0 - formatted | 20.8 | 8.2 |
|