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 19, 2024

Commit

297a806

·

verified ·

1 Parent(s): 6df2bf6

Update README.md

Files changed (1) hide show

README.md +1 -1

README.md CHANGED Viewed

@@ -19,7 +19,7 @@ This repository contains evaluation results from the Malayalam ASR model "vrclc/
 - **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%.

 - **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(%): 49.67724937425899