CAPR — Computer-Assisted Proto-language Reconstruction (Burmish 2.0)

This repository contains a re-write and working implementation of the CAPR interface and backend for reconstructing Proto‑Burmish (Lolo‑Burmese branch). It accompanies the Zenodo record:

• Knights, Seth; Gong, Xun; Hill, Nathan W. (2022). Materials for an Etymological Dictionary of Burmish 2.0. Zenodo. DOI: 10.5281/zenodo.6617027

What’s here (high level)

• server/ — Python/Flask API for CAPR
• cognate-app/ — Svelte interface for cognate reassignment and FST editing
• pipeline/ — Wordlist → tokenized lexicon → initial cognate discovery (LingPy/LingRex)
• reconstruct/ — Initial FSTs used by the pipeline
• orthoprofiles/ — Orthographic profiles for pipeline stages

Notes on setup and run below paraphrase the Zenodo page for 10.5281/zenodo.6617027.

Quick start (development)

1. API

   cd server
   export FLASK_APP=server
   flask run    # API on http://localhost:5000

2. Interface

   cd cognate-app
   npm i
   npm run dev  # UI on http://localhost:8080

Foma / FST toolchain

Install libfoma0 and libfoma0-dev (Linux). Some distro versions (e.g., 0.9.18) can yield uncaught FST compilation errors. If that affects you, build a recent version from source.

Linux build from source (example):

wget https://github.com/mhulden/foma/archive/refs/heads/master.zip
unzip master.zip
cd foma-master/foma
make
sudo make install

Language → .bin mapping (required in the FST editor)

Old_Burmese → burmese.bin
Achang_Longchuan → ngochang.bin
Xiandao → xiandao.bin
Maru → maru.bin
Bola → bola.bin
Atsi → atsi.bin
Lashi → lashi.bin

Production sketch

• API: gunicorn server:app (e.g., :8000)
• UI: npm run build && npm run start (:8080)
• Reverse proxy: e.g., Caddy

Cite

If you use this software or datasets, please cite:

• Knights, S.; Gong, X.; Hill, N. W. (2022). Materials for an Etymological Dictionary of Burmish 2.0. Zenodo. https://doi.org/10.5281/zenodo.6617027

Authors

• Seth Knights — University of Chicago
• Xun Gong — University of Vienna
• Nathan W. Hill — Trinity College Dublin

License

Zenodo lists “Other (Public Domain)”. This repo uses CC0‑1.0 to express a public‑domain dedication. See LICENSE.