Eliezermga/Lugayetu
Lugayetu is a research and technology project focused on preserving Congolese low-resource languages and developing artificial intelligence tools for them. The platform collects textual and voice data from native speakers to build ai tools (translator...)
What's novel
Lugayetu is a research and technology project focused on preserving Congolese low-resource languages and developing artificial intelligence tools for them. The platform collects textual and voice data from native speakers to build ai tools (translator...)
Code Analysis
20 files read · 4 roundsA Django web platform for collecting Ruund-French parallel linguistic data (text and audio) from native speakers, serving a pre-trained mBART model via a REST API for Ruund↔French translation, with a corpus-building pipeline that extracts, aligns, and deduplicates Bible translations.
Strengths
The project has a coherent end-to-end pipeline: PDF extraction → verse alignment → parallel corpus creation → mBART inference API → web-based data collection with moderation workflow. The mBART integration with custom language token handling (ruu_CM) and the DRC-specific user model with province/mother-tongue tracking show genuine domain-specific thinking.
Weaknesses
Test coverage is essentially zero (all test files are ~60 bytes of boilerplate). The corpus scripts (main.py, work/main.py) are near-duplicate copies, the PDF extractor auto-installs packages at runtime (fragile), and the translation model is a thin inference wrapper with no training, fine-tuning, or evaluation code in the repository. The README overstates the project as an 'AI research platform' when the actual ML component is a single Hugging Face model call.
Score Breakdown
Signal breakdown
Innovation
Craft
Traction
Scope
Evidence
Commits
73
Contributors
3
Files
91
Active weeks
9
Repository
Language
Python
Stars
2
Forks
0
License
MIT