Reproduce “GFLean: An Autoformalisation Framework for Lean via GF” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Shashank Pathak
GFLean is an autoformalisation framework that translates mathematical statements from a controlled natural language, Simplified ForTheL, into formal Lean expressions. It successfully parses and formalizes 42 out of 62 statements from a mathematics textbook chapter, demonstrating a rule-based approach using Grammatical Framework (GF) for linguistic processing and Haskell for Abstract Syntax Tree transformations.
Key insights
- Grammatical Framework (GF) provides a modular and potentially scalable approach for building rule-based autoformalisation programs, effectively separating abstract and concrete syntax.
- The combination of GF for parsing and linearization with Haskell for complex AST transformations offers an effective pipeline for translating controlled natural language to formal expressions.
- Rule-based autoformalisation systems can serve as robust components, potentially complementing neural network approaches in future hybrid autoformalisation systems.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.