About Zomi‑Syl

Zomi‑Syl is a modular, dialect‑aware syllabification engine designed for the Zomi language and its closely related varieties.
It provides a unified, linguistically grounded framework for syllable segmentation across dialects, orthographies, and backend architectures.

Zomi‑Syl is built for:

  • linguists
  • NLP researchers
  • educators
  • community contributors
  • developers integrating Zomi processing into applications

It is open‑source, reproducible, and designed for long‑term maintainability.

🎯 Core Goals

Zomi‑Syl was created to solve several long‑standing challenges in Zomi language technology:

  • inconsistent syllabification across dialects
  • lack of standardized tooling
  • absence of reproducible, testable backends
  • difficulty integrating rule‑based and ML‑based approaches
  • need for a single, authoritative pipeline for research and production

The project aims to provide:

  • dialect‑aware syllabification
  • multiple interchangeable backends
  • transparent evaluation and benchmarking
  • clean API + CLI
  • a unified metadata schema
  • a single source of truth for profiles and rules

🧩 Architecture Overview

Zomi‑Syl is built on a modular pipeline:

  • Profiles — dialect‑specific phonotactic inventories
  • Backends — rule‑based, CRF, transformer, and future models
  • Registry — unified backend and profile discovery
  • Evaluation — benchmarking, metrics, reports
  • CLI — user‑friendly command‑line interface
  • Python API — programmatic access

This design ensures:

  • reproducibility
  • extensibility
  • dialect flexibility
  • backend neutrality
  • clean separation of concerns

🌏 Dialect‑Aware by Design

Zomi‑Syl supports multiple Zomi varieties, including:

Language / Group ISO‑639‑3 Link Glottolog Link
Gangte gnb gang1266
Kom Rem kmm komr1235
Mate
Paite pck pait1246
Simte smt simt1235
Siyin csy siyi1235
Tedim ctd tedi1235
Thado / Thadou tcz thad1235
Thangkhal
Vaiphei vap vaip1236
Zo / Zou zom zouu1234
India Zomi
Myanmar Zomi
Zolai Standard

Each dialect has its own:

  • onset inventory
  • nucleus inventory
  • coda inventory
  • tone behavior (if applicable)
  • orthographic conventions
  • rule set

See Dialect Profiles for details.

🛠 Backends

Zomi‑Syl supports multiple interchangeable backends:

  • Rule‑based
  • CRF (Conditional Random Field)
  • Transformer‑based (experimental)
  • Reverse rule‑based (diagnostic)

Each backend is:

  • versioned
  • benchmarked
  • validated
  • documented

See Backends for implementation details.

📊 Evaluation & Benchmarking

Zomi‑Syl includes a full evaluation suite:

  • accuracy
  • boundary F1
  • confusion matrices
  • confidence scoring
  • regression testing
  • dataset validation

See Benchmarking Guide.

📦 Installation & Usage

  • Install: Installation
  • Try the CLI: CLI Reference
  • Try the API: Quick Start

🤝 Community & Contributions

Zomi‑Syl is a community‑driven project.
Contributions are welcome from:

  • linguists
  • developers
  • educators
  • native speakers
  • researchers

See:

  • Linguistic Contributors Guide
  • Developer Guide
  • Infra Maintainers Guide

📜 License

Zomi‑Syl is released under the MIT License, allowing:

  • academic use
  • commercial use
  • modification
  • redistribution

Next Steps

  • Explore dialects: Dialect Profiles
  • Learn the CLI: CLI Reference
  • Add a backend: Adding New Backends
  • Benchmark models: Benchmarking Guide