Datasets:

nectec
/

LOTUSDIS

license: cc-by-sa-4.0
language:
  - th
tags:
  - speech-recognition
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
      - split: test
        path: data/test-*
dataset_info:
  features:
    - name: audio
      dtype:
        audio:
          sampling_rate: 16000
    - name: sentence
      dtype: string
    - name: speaker_id
      dtype: string
    - name: mic
      dtype: string
    - name: duration
      dtype: float64
  splits:
    - name: train
      num_bytes: 8212128894.78
      num_examples: 120245
    - name: validation
      num_bytes: 1296622162.01
      num_examples: 13090
    - name: test
      num_bytes: 1623791447.32
      num_examples: 27580
  download_size: 13180732521
  dataset_size: 11132542504.109999

LOTUSDIS

Dataset Description

How to use

You can easily load the dataset using the 🤗 datasets library. The dataset can be loaded and prepared with a single line of Python code:

from datasets import load_dataset
lotus_dis = load_dataset("nectec/LOTUSDIS", split="train")

To iterate through the dataset without downloading it entirely, you can use streaming mode:

from datasets import load_dataset
lotus_dis = load_dataset("nectec/LOTUSDIS", split="train", streaming=True)
print(next(iter(lotus_dis)))

Learn more about how to load and prepare audio datasets in the Hugging Face Audio Datasets tutorial.

Full meeting session resources:

Audio files: Download here
Annotation files (TextGrid): Download here

Citation

@misc{tipaksorn2025lotusdisthaifarfieldmeeting,
      title={LOTUSDIS: A Thai far-field meeting corpus for robust conversational ASR}, 
      author={Pattara Tipaksorn and Sumonmas Thatphithakkul and Vataya Chunwijitra and Kwanchiva Thangthai},
      year={2025},
      eprint={2509.18722},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2509.18722}, 
}