Open-Labor-PH
About the Project
Open-Labor-PH is open-source infrastructure for making Philippine labor law usable by AI systems. Decades of DOLE Orders and regulations live as unstructured PDFs, which makes them nearly impossible to query reliably or ground a language model against. This project converts that corpus into validated, machine-readable JSON and pairs it with a retrieval-augmented-generation (RAG) evaluation suite so retrieval quality over the dataset can be measured rather than guessed at. The ingestion and benchmarking pipeline is written in Rust (serde) for performance, with Python interop through PyO3 for ML and evaluation workflows.
Key Features
- Decades of DOLE Orders converted from PDF into structured JSON
- JSON Schema validation to keep the dataset consistent and trustworthy
- RAG evaluation suite for benchmarking retrieval over labor law
- High-performance Rust ingestion pipeline (serde)
- Python interop via PyO3 for ML/eval workflows
- Open data foundation others can build legal-tech tools on
Challenges & Solutions
Legal text is messy: PDFs with inconsistent formatting, amendments that modify earlier orders, and edge cases that resist clean schematization. Designing a JSON Schema that faithfully captures the structure of labor regulations — and a validation pass that catches bad extractions — is as much a domain problem as an engineering one. On top of that, building an RAG evaluation suite means defining what 'correct retrieval' even means for legal questions.
Outcomes & Impact
Open-Labor-PH is open source and intended as shared infrastructure: a structured, validated dataset and evaluation harness that other developers can build Philippine legal-tech and compliance tools on, rather than each re-scraping the same PDFs.
Interested in Open-Labor-PH?
Let's collaborate or discuss how I can help bring your projects to life.
Get in touch