Dharmamitra Parallel Aligned Text Corpus — Website

An interactive HTML corpus browser for the MITRA-parallel dataset, covering sentence-aligned parallel texts in Sanskrit, Tibetan, and Chinese.

Features

• Foldable navigation tree: language pair → collection → category → text
• All 6 alignment directions: SA↔BO, SA↔ZH, BO↔ZH
• Inverted views: each TSV file is accessible from both directions
• Sentence-pair display with language-appropriate fonts
• Search across text names

Setup

1. Build the navigation data

Run the build script from this directory:

python3 build.py

This reads metadata from metadata/ and TSV filenames from ../mitra-parallel/mitra-parallel/tsv/, and generates docs/nav_data.js. Requires the mitra-parallel repo to be checked out as a sibling:

code/
  mitra-parallel/          ← https://github.com/dharmamitra/mitra-parallel
  mitra-parallel-website/  ← this repo

2. Run locally

Create a symlink so the local server can serve TSV files:

ln -s /path/to/mitra-parallel/mitra-parallel/tsv docs/tsv

Then start a local HTTP server (the page must be served via HTTP, not opened as a file):

cd docs && python3 -m http.server 8765
# Open http://localhost:8765/

The site auto-detects localhost and uses ./tsv/ as the data URL.

3. Deploy to GitHub Pages

1. Push this repo to GitHub
2. In repo Settings → Pages, set source to Deploy from branch, branch main, folder /docs
3. The site will auto-use https://raw.githubusercontent.com/dharmamitra/mitra-parallel/main/mitra-parallel/tsv/ as the TSV source

You can also override the TSV URL at runtime via the yellow bar at the top of the page.

Files

docs/
  index.html     — main website (self-contained HTML/CSS/JS)
  nav_data.js    — generated navigation index (~4MB, ~22k entries)
  tsv/           — symlink or copy of mitra-parallel/mitra-parallel/tsv/
metadata/
  SA_files.json, SA_category-names.json, SA_collection-names.json
  BO_files.json, BO_category-names.json, BO_collection-names.json
  ZH_files.json, ZH_category-names.json, ZH_collection-names.json
build.py         — navigation builder

Data

The alignment data comes from the MITRA-parallel corpus:

• 12,481 TSV files covering SA-BO, SA-ZH, and BO-ZH pairs
• ~90% of files matched to metadata with display names
• Released under CC BY-SA 4.0

Citation: Sebastian Nehrdich & Kurt Keutzer, MITRA: A Large-Scale Parallel Corpus and Multilingual Pretrained Language Model for Machine Translation and Semantic Retrieval, 2026.