CoolFace
Datasetpublic

malaysia-ai/Malaysian-STT

Malaysian-STT Prepare streaming and whole mode Speech-to-Text Malaysian context dataset, suitable to train streaming LLM base or Encoder-Decoder such as Whisper. Merged 30 seconds chunk into one audio file, can up to 10 minutes. Segmentize based on silent at least 0.3 seconds. Reject low score based on force alignment. Reject timestamp anomaly based on force alignment. Dataset involved Dialects IMDA Malaysian context Malaysia Parliament Science context… See the full description on the dataset page: https://huggingface.co/datasets/malaysia-ai/Malaysian-STT.

sourceHugging Faceupdated 1y agoView on Hugging Face
2likes1.3kdownloads
Dataset Card

Malaysian-STT

Prepare streaming and whole mode Speech-to-Text Malaysian context dataset, suitable to train streaming LLM base or Encoder-Decoder such as Whisper.

  1. 1.Merged 30 seconds chunk into one audio file, can up to 10 minutes.
  2. 2.Segmentize based on silent at least 0.3 seconds.
  3. 3.Reject low score based on force alignment.
  4. 4.Reject timestamp anomaly based on force alignment.

Dataset involved

  1. 1.Dialects
  2. 2.IMDA
  3. 3.Malaysian context
  4. 4.Malaysia Parliament
  5. 5.Science context
  6. 6.Synthetic

how to prepare the dataset

bash
huggingface-cli download --repo-type dataset \
--include '*segment*.zip' \
--local-dir './' \
--max-workers 20 \
malaysia-ai/Malaysian-STT

huggingface-cli download --repo-type dataset \
--include '*whole*.zip' \
--local-dir './' \
--max-workers 20 \
malaysia-ai/Malaysian-STT

wget https://gist.githubusercontent.com/huseinzol05/2e26de4f3b29d99e993b349864ab6c10/raw/9b2251f3ff958770215d70c8d82d311f82791b78/unzip.py
python3 unzip.py

Speech tokens

We also converted the audio to speech tokens using GLM4 Audio Tokenizer,

bash
huggingface-cli download --repo-type dataset \
--include 'glm4-*.zip' \
--local-dir './' \
--max-workers 20 \
malaysia-ai/Malaysian-STT

wget https://gist.githubusercontent.com/huseinzol05/2e26de4f3b29d99e993b349864ab6c10/raw/9b2251f3ff958770215d70c8d82d311f82791b78/unzip.py
python3 unzip.py

With total 3283240919 speech tokens equivalent to 72960.9 hours.

Source code

  1. 1.Source code for data processing at https://github.com/malaysia-ai/dataset/tree/main/streaming-stt
  2. 2.Source code for audio tokenizer at https://github.com/malaysia-ai/cooking/tree/main/qwen-stt

Acknowledgement

Special thanks to Lambda Research Grant program for Lambda cloud credit!