Tedim Zolai language technology & AI research
Zolai AI builds trustworthy language data, knowledge infrastructure, NLP tools, literacy resources, and community-oriented applications for Tedim Zolai (ZVS 2018) — with specialized model training as a later stage when the foundation is reliable.
Create practical, evidence-based language technology and knowledge infrastructure that help Tedim Zolai speakers and learners access information, learn language, and participate digitally.
A durable Zomi language-technology ecosystem: community-informed tools, open research, measurable literacy impact, and sustainable operations.
Status labels: Implemented · Experimental · Planned · Live
| Category | Focus | Status |
|---|---|---|
| Language data | Dictionary, Bible parallel corpus (where permitted), grammar, syllables | Implemented (ongoing quality) |
| NLP | Syllable segmentation strong; POS, morphology, MT, etc. | Mixed — see notes |
| Applications | Web learner, desktop (early), MCP for AI assistants | Live / Early |
| Education | Vocab, curriculum, exercises | Experimental → Planned |
| Research | Low-resource NLP, evaluation, community validation | Planned / in progress |
NLP honesty note: Syllable tooling has the strongest evidence. Broader NLP modules exist in code but many claims remain only partially verified — see claims audit.
We credit source materials used to build the Zolai learning system. Source corpora
(Bible translations, TongDot/TongSan dictionaries, and other corpora) are processed into a
cleaned, ZVS-2018-aligned database (data/zolai.db, SQLite WAL, ~99 tables / ~2.3GB / ~3.3M rows).
See zolai-datasets CREDITS for attribution. Public accessibility ≠ unrestricted reuse.
| Resource | Entries | Purpose |
|---|---|---|
dictionary (ZO→EN) |
84,490 | Cleaned master Zolai→English |
dictionary_en_zo (EN→ZO) |
64,025 | English→Zolai (+ related fields) |
bible_verses |
31,102 | Parallel EN/ZO/MY verses |
translations |
207,623 | Sentence pairs |
syllable_data |
189,563 | Syllable segmentation |
training_exercises |
82,159 | Exercise types |
| Service | URL |
|---|---|
| Landing | https://zolai.space/ |
| MCP Server | https://mcp.zolai.space/mcp |
| Documentation | https://github.com/Zolai-AI/.github/tree/main/docs |
| Repo | Purpose | Status |
|---|---|---|
zolai-core |
Python toolkit + RAG Knowledge Brain | Active |
zolai-web |
Learner platform (Next.js + Hono + Prisma) | Active |
zolai-tauri |
Offline desktop app (Tauri 2) | Early |
zolai-datasets |
Bilingual corpora & datasets | Active |
zolai-training |
LoRA/QLoRA + GGUF (later-stage focus) | Early |
zolai-wiki |
Knowledge base (grammar, vocab, curriculum) | Active |
zolai-mcp-server |
Legacy MCP wrapper — the live MCP server is a route in zolai-core |
Legacy |
zolai-landing |
Org landing (zolai.space) | Live |
Parallel Bible text is a major EN/ZO resource used as a language learning / research corpus, not for religious advocacy. License and permission status must be documented before redistribution or commercial reuse.
| Role | Name | Notes |
|---|---|---|
| Founder & Technical Lead | Peter Pau Sian Lian (@peterlianpi) |
Architecture, engineering, execution |
| Whitepaper Ideas Contributor | Shwe Yee | Ideas for whitepaper only — advisor role not confirmed |
| Document | Description |
|---|---|
| Project context | Canonical context for agents & contributors |
| Mission & vision | Identity and pillars |
| Source of truth | Canonical doc owners |
| OKRs | Annual + 90-day objectives |
| Strategic audit | Full ecosystem audit |
| Docs index | Master documentation index |
- Open source code — application code MIT-licensed where stated per-repo; data is not blanket-open (Bible/dictionary sources often RESTRICTED — see
docs/governance/credits-license-inventory.md) - Evidence first — data, evaluation, and community impact before model hype
- Community-aware — Zomi speakers as stakeholders, not only data sources
- ZVS 2018 — orthography ground truth; SOV; ergative
in - AI-assisted development — human-directed; see AI Tools & Attribution
| Role | How to help |
|---|---|
| Zomi speaker | Validate translations, test tools, contribute cultural knowledge |
| Linguist | Grammar, dialect documentation, morphology |
| ML engineer | NLP pipelines, evaluation, careful training experiments |
| Developer | Web / desktop / API / infrastructure |
| Educator | Curriculum, exercises, pedagogical grammar |
See CONTRIBUTING.md.
- GitHub Issues — bugs, features
- GitHub Discussions — questions, ideas
- Email — peterpausianlian2020@gmail.com
Lungdam! (Thank you!)
Built for the Zomi people
Led by Peter Pau Sian Lian (@peterlianpi) · AI-assisted with Cursor + OpenCode