minhtien2405/fptu-vovinam-dataset

FPTU Vovinam Dataset

Dataset Description

This dataset contains Vietnamese audio recordings with corresponding text transcriptions, specifically focused on Vovinam martial arts terminology and instruction. This dataset is created by FPT University for research and educational purposes in the field of Vietnamese speech recognition and natural language processing.

Dataset Structure

• —Audio Files: High-quality audio recordings in AAC format (uploaded as raw data)
• —Text: Vietnamese transcriptions of the audio content
• —Metadata: Speaker information, environment, gender, location data
• —Splits: Train (80%), Validation (10%), and Test (10%) sets

Dataset Features

Usage

Loading the Dataset

from datasets import load_dataset

# Load the dataset
dataset = load_dataset("minhtien2405/fptu-vovinam-dataset")

# Access different splits
train_data = dataset['train']
val_data = dataset['validation']
test_data = dataset['test']

# Get a sample
sample = train_data[0]
audio = sample['audio']
text = sample['text']

Working with Audio

# Access audio data
audio_array = audio['array']
sampling_rate = audio['sampling_rate']

# Play audio (if in Jupyter)
import IPython.display as ipd
ipd.Audio(audio_array, rate=sampling_rate)

# Convert to numpy array for processing
import numpy as np
audio_np = np.array(audio_array)

Text Processing

# Access text
text = sample['text']
print(f"Transcription: {text}")

# Get metadata
speaker_id = sample['speakerId']
environment = sample['environment']
gender = sample['gender']
province = sample['province_name']
region = sample['region']

Dataset Statistics

• —Total samples: 17,560
• —Audio format: AAC (stored as raw data)
• —Language: Vietnamese
• —Domain: Vovinam martial arts
• —Institution: FPT University
• —Recording quality: High-quality audio recordings
• —Geographic coverage: Multiple provinces across Vietnam

Data Splits

Use Cases

This dataset is suitable for:

• —Automatic Speech Recognition (ASR): Training Vietnamese ASR models with regional accent variations.
• —Text-to-Speech (TTS): Building Vietnamese TTS systems, especially for Vovinam terminology and commands.
• —Speech Analysis: Studying pronunciation, intonation, and regional accent differences in martial arts contexts.
• —Martial Arts Education: Developing learning and training applications for FPTU students and the wider Vovinam community.
• —Language Research: Supporting studies in Vietnamese linguistics, dialects, and phonetics within martial arts education.

Technical Details

• —Audio Format: AAC (Advanced Audio Coding)
• —Sampling Rate: Preserved from original recordings (variable)
• —Channels: Mono/Stereo (preserved from original)
• —Bit Depth: Preserved from original recordings
• —File Size: Optimized for web delivery

Citation

If you use this dataset in your research, please cite:

@dataset{fptu_vovinam_dataset,
  title={FPTU Vovinam Dataset: Vietnamese Speech Recognition for Martial Arts Education},
  author={FPT University},
  year={2025},
  url={https://huggingface.co/datasets/minhtien2405/fptu-vovinam-dataset},
  note={A Vietnamese speech dataset focusing on regional accents and Vovinam martial arts terminology}
}

Contact

For questions or issues regarding this dataset, please contact:

• —Email: phamminhtien2405@gmail.com
• —Repository: https://huggingface.co/datasets/minhtien2405/fptu-vovinam-dataset

Acknowledgments

We sincerely thank the students and Vovinam practitioners at FPTU, along with the wider community, for their contributions in creating this valuable dataset for Vietnamese speech recognition research in martial arts.

This dataset is part of FPT University's ongoing research in Vietnamese natural language processing and speech recognition technologies.