Datasets:

asr-malayalam
/

Mal_ASR_Predict_Ref_Samples

Tasks:

Sentence Similarity

Modalities:

Formats:

Languages:

Size:

Libraries:

License:

Dataset card Data Studio Files Files and versions

kavyamanohar commited on Jul 11, 2024

Commit

96d69de

·

verified ·

1 Parent(s): a09012a

Update README.md

Files changed (1) hide show

README.md +3 -2

README.md CHANGED Viewed

@@ -16,13 +16,14 @@ This repository contains evaluation results from the Malayalam ASR model "vrclc/
 * Dataset: [google/fleurs](https://huggingface.co/datasets/google/fleurs/viewer/ml_in/test)
-### Dataset Description
 - **Curated by:** VRCLC
 - [vrclc/Whisper_small_malayalam](https://huggingface.co/vrclc/Whisper_small_malayalam) was trained with 50 hours of Malayalam speech data.
 - The test set of [google/fleurs](https://huggingface.co/datasets/google/fleurs/viewer/ml_in/test) dataset which consists of Malayalam speech data was used to evaluate the model
 - The evaluation of 500 samples from [google/fleurs](https://huggingface.co/datasets/google/fleurs/viewer/ml_in/test) and gives an unnormalized WER is 60.4%.
 ## Uses

 * Dataset: [google/fleurs](https://huggingface.co/datasets/google/fleurs/viewer/ml_in/test)
 - **Curated by:** VRCLC
 - [vrclc/Whisper_small_malayalam](https://huggingface.co/vrclc/Whisper_small_malayalam) was trained with 50 hours of Malayalam speech data.
 - The test set of [google/fleurs](https://huggingface.co/datasets/google/fleurs/viewer/ml_in/test) dataset which consists of Malayalam speech data was used to evaluate the model
 - The evaluation of 500 samples from [google/fleurs](https://huggingface.co/datasets/google/fleurs/viewer/ml_in/test) and gives an unnormalized WER is 60.4%.
 ## Uses