undertheseanlp/uts2025_vietipa

Vietnamese IPA Dataset

A comprehensive Vietnamese IPA (International Phonetic Alphabet) dataset with word pronunciations and MP3 audio files for text-to-speech and pronunciation learning applications.

Dataset Description

Dataset Summary

This dataset contains 50 common Vietnamese words with their IPA (International Phonetic Alphabet) transcriptions and corresponding audio files. It's designed for:

• Text-to-speech systems development
• Vietnamese pronunciation learning
• Phonetic research
• Language learning applications

Supported Tasks

• Text-to-Speech: Generate Vietnamese speech from text
• Pronunciation Learning: Learn correct Vietnamese pronunciation with IPA
• Phonetic Analysis: Study Vietnamese phonetics with accurate IPA transcriptions

Languages

Vietnamese (vi)

Dataset Structure

Data Instances

Each instance contains:

• A Vietnamese word or phrase
• Its IPA transcription with tone marks
• Path to the corresponding MP3 audio file

Data Fields

• file_name (Audio): Audio file path that will be automatically loaded by Hugging Face Datasets
• word (string): Vietnamese word or phrase
• ipa (string): IPA transcription with tone marks

Data Splits

The dataset contains a single training split with 50 examples.

Dataset Creation

Curation Rationale

This dataset was created to provide a simple, accessible resource for Vietnamese pronunciation with accurate IPA transcriptions and audio, addressing the lack of small-scale Vietnamese phonetic datasets.

Source Data

Initial Data Collection

Words were selected from common Vietnamese vocabulary including:

• Greetings and basic phrases
• Family terms
• Numbers
• Time expressions
• Common nouns and verbs

Annotations

Annotation Process

IPA transcriptions were created following standard Vietnamese phonetic conventions, including tone marks:

• ˧˧ (33): Level tone (ngang)
• ˨˩ (21): Falling tone (huyền)
• ˧˩ˀ (32ˀ): Falling-rising tone (hỏi)
• ˧˥ (35): Rising tone (sắc)
• ˧˩˧ (313): Broken tone (ngã)
• ˨˩ˀ (21ˀ): Drop tone (nặng)

Audio files were generated using Google Text-to-Speech (gTTS) for consistency.

Usage

Quick Start

from datasets import load_dataset

# Load the dataset from Hugging Face Hub
dataset = load_dataset("undertheseanlp/uts2025_vietipa")

# Access the data
for example in dataset['train']:
    word = example['word']
    ipa = example['ipa']
    audio = example['file_name']  # Audio object with array, path, and sampling_rate
    print(f"{word}: {ipa}")
    print(f"Audio path: {audio['path']}")
    print(f"Audio sampling rate: {audio['sampling_rate']} Hz")

Audio Processing Example

from datasets import load_dataset

# Load the dataset
dataset = load_dataset("undertheseanlp/uts2025_vietipa")

# Process audio data
for example in dataset['train'].select(range(3)):  # First 3 examples
    word = example['word']
    ipa = example['ipa']
    audio = example['file_name']

    print(f"Word: {word}")
    print(f"IPA: {ipa}")
    print(f"Audio shape: {audio['array'].shape}")
    print(f"Sampling rate: {audio['sampling_rate']} Hz")
    print("-" * 40)

Local Development

This project uses uv for dependency management:

# Install dependencies
uv sync

# Generate audio files
uv run python generate_audio.py generate

# Display dataset
uv run python generate_audio.py display

# Add new word
uv run python generate_audio.py add "xin lỗi" "sin˧˧ loj˧˩"

# Interactive example
uv run python example_usage.py

Example Data

Sample JSONL Records

{"file_name": "audio/xin_chao.mp3", "word": "xin chào", "ipa": "sin˧˧ caːw˨˩"}
{"file_name": "audio/cam_on.mp3", "word": "cảm ơn", "ipa": "kaːm˧˩ ʔəːn˧˧"}
{"file_name": "audio/viet_nam.mp3", "word": "Việt Nam", "ipa": "viət˧˩ˀ naːm˧˧"}
{"file_name": "audio/nuoc.mp3", "word": "nước", "ipa": "nɨək˧˥"}
{"file_name": "audio/yeu.mp3", "word": "yêu", "ipa": "iəw˧˧"}

Data Table View

Audio File	Word	IPA
audio/xin_chao.mp3	xin chào	sin˧˧ caːw˨˩
audio/cam_on.mp3	cảm ơn	kaːm˧˩ ʔəːn˧˧
audio/viet_nam.mp3	Việt Nam	viət˧˩ˀ naːm˧˧
audio/nuoc.mp3	nước	nɨək˧˥
audio/yeu.mp3	yêu	iəw˧˧

Considerations for Using the Data

Social Impact

This dataset promotes Vietnamese language learning and preservation through accurate phonetic representations.

Limitations

• Limited to 50 common words (suitable for basic applications)
• Audio generated via TTS, not native speaker recordings
• IPA transcriptions follow Northern Vietnamese dialect
• May not capture all regional pronunciation variations

Recommendations

• For production systems, consider native speaker recordings
• Expand vocabulary for comprehensive coverage
• Add regional dialect variations if needed

Additional Information

Dataset Curators

Created for educational and research purposes in Vietnamese phonetics and pronunciation.

IPA Package

Tipa package for LaTeX is used for IPA symbols:

• CTAN link: https://ctan.org/pkg/tipa
• Documentation: https://sg.mirrors.cicku.me/ctan/fonts/tipa/tipa/doc/tipaman.pdf

Licensing Information

This dataset is released under the Apache 2.0 License.

Citation Information

If you use this dataset, please cite:

@dataset{vietnamese_ipa_2025,
  author = {Trang Phan, Vu Anh},
  organization = {Underthesea NLP}
  title = {Vietnamese IPA Dataset},
  year = {2025},
  publisher = {Hugging Face},
  url = {https://huggingface.co/datasets/undertheseanlp/uts2025_vietipa}
}

Contributions

Contributions are welcome! Please feel free to:

• Add more words with IPA transcriptions
• Improve existing IPA transcriptions
• Add native speaker recordings
• Include regional dialect variations