laolengsaeu/shan-phonology-master-database

Shan Language Acoustic Phonology Master Database

Project Overview

This dataset represents the world’s first systematic, comprehensive acoustic mapping of native Shan language (လိၵ်ႈတႆး / Tai Shan) phonology explicitly designed for generative AI, Text-to-Speech (TTS), and Artificial Intelligence Singing Voice Synthesis (SVS) engines.

Historically, minority and regional languages under the Tai-Kadai group suffer from being heavily "under-resourced" in the digital domain. This project bridges the digital divide by isolating every valid phonetic cell in the Shan writing system across the standardized Six-Tone Contrast System, completely mapped from an indigenous linguistic perspective.

Creator & Intellectual Property

• Principal Investigator & Native Speaker: Sai Lao Leng Saeu (ကိုစိုင်းလောဝ်)
• Dataset Architecture: Co-developed with AI Systems Integration (May 2026)
• Status: 100% Verified Foundational Core Locked

---

Dataset Structure & Specifications

The dataset strictly follows the standardized structural engineering and construction hierarchy of the Shan syllable matrix. Instead of continuous conversational speech, which introduces co-articulation blur, this database consists of cleanly isolated, high-fidelity tokens categorized into 5 distinct computational phases:

1. Phase 1: မႄႈၵပ်းငဝ်ႈ (Core Vowel Systems) — Mapping of the 10 foundational vowel matrices across all 6 tones.
2. Phase 2: မႄႈၵပ်းသွၼ်ႉ (Diphthongs) — Secondary vowel systems (including ဢႆ၊ ဢဝ်၊ ဢႂ် structures) mapped entirely across the 6-tone spectrum.
3. Phase 3: မႄႈသဵင်သွၼ်ႉ (Medial Consonant Clusters) — Consonant glide couplings utilizing medials: ယ (ႁွပ်ႈ), ရ (လဵပ်ႈ), and ဝ (ၵႂၢႆႉ) across all tone levels.
4. Phase 4: တူဝ်ၽႅတ်းသႅင်လင် (Nasal Finals / Continuous Sounds) — Closed syllables ending in continuous nasal tracts: -မ် (Bilabial), -ၼ် (Alveolar), and -င် (Velar) finals, mapped for pitch length variations.
5. Phase 5: တူဝ်ၽႅတ်းသဵင်ၶၢတ်ႇ (Stop Finals / Abrupt Breaks) — Closed syllables ending in unreleased sudden stops: -ပ် (Bilabial Stop), -တ် (Alveolar Stop), and -ၵ် (Velar Stop), mapping sharp pitch breaks.

Audio Properties

• Acoustic Fidelity: Studio-clean, dry-vocal profile optimized for deep learning models.
• Tonal Verification: 100% of recorded tokens manually verified via digital pitch tracking to prevent tonal drift or vocal slurring.

---

Licensing & Terms of Use

This repository is strictly protected under the Creative Commons Attribution-NonCommercial 4.0 International License (CC BY-NC 4.0).

What You Can Do:

• Share & Adapt: You may copy, redistribute, extract, and build upon this data for academic research, personal projects, or open-source linguistic software developments.

What You Must Comply With:

• Attribution (BY): You must give appropriate credit to the creator (Sai Lao Leng Saeu), provide a link to this Huging Face repository, and indicate if changes were made.
• NonCommercial (NC): You cannot use this dataset, or any models trained directly on this data (such as TTS engines, RVC voice clones, or singing tools), for commercial profit or monetization without written consent from the author.

For academic citations, please credit this repository as:

Sai Lao Leng Saeu (2026). Shan Language Acoustic Phonology Master Database for Generative AI Systems. Hugging Face Dataset Repository.