📇 Dataset Card: meeting_transcription_audio_th

Dataset Summary

A Thai‐language speech corpus of online meeting recordings paired with gold‐standard transcripts, originally compiled for the 2025-ASR Automatic Speech Recognition Task a public competition organized by AI Thailand Benchmark Programs on Benchmark.ai (started May 25 2025) to push the state of the art in robust, multi-speaker ASR for Thai. The corpus includes both real meeting audio and acoustically augmented variants to support research on noise, reverberation, and overlapping-speaker scenarios. I have since collected, standardized, and converted these recordings and transcripts into a Hugging Face–compatible dataset, providing preprocessed audio files, enriched metadata, and ready-to-load scripts to streamline experimentation and integration on the HF platform.

Supported Tasks

• Automatic Speech Recognition (ASR)
• Multi-speaker speech transcription
• Robust ASR under adverse acoustic conditions

Languages

• Thai (th)

---

Dataset Structure

Each example in the dataset is a dictionary with the following fields:

Field	Type	Description
audio	audio	The raw audio file (e.g. WAV, 16 kHz)
transcription	string	The verbatim Thai transcript (gold standard)

Splits

Split	Sessions	Total Duration	Notes
train	80 sessions	~ 20 hours	Real online-meeting recordings with accurate transcripts.
validation	10 sessions	~ 2.5 hours	Held-out sessions for model tuning and hyperparameter selection.
test	40 sessions (10 real + 30 aug.)	~ 10 hours	10 real + 30 artificially augmented (noise/reverb/overlap); transcripts withheld for blind evaluation.

---

Usage

from datasets import load_dataset

dataset = load_dataset("Konthee/meeting_transcription_audio_th")
train_split = dataset["train"]
audio_sample = train_split[0]["audio"]
transcript   = train_split[0]["transcription"]

Example

{
  "audio": audio,
  "transcription": "เพื่อน ได้ยิน หรือเปล่า"
}

APA

AI Thailand Benchmark Programs. (2025). 2025-ASR: Automatic Speech Recognition Task. Retrieved June 23, 2025, from https://benchmark.ai.in.th/task/detail/2025-asr

---

License

@misc{meeting_transcription_audio_2025,
  title        = {meeting_transcription_audio_th: A Thai Online-Meeting Speech Corpus for Multi-Speaker ASR},
  author       = {AI Thailand Benchmark Programs and Konthee Bo},
  year         = {2025},
  howpublished = {https://huggingface.co/datasets/Konthee/meeting_transcription_audio_th},
  note         = {Dataset reformatted and packaged by Konthee Bo; original data from the 2025-ASR competition},
  license      = {CC-BY-SA 4.0}
}

Authors

• Konthee Boonmeeprakob ([email protected])
• Pitikorn Khlaisamniang ([email protected])