Knowledge graphs, but make them hyperedges

What it does

Hyperbase is a Python library for building Semantic Hypergraphs (SH). Instead of flattening a sentence into subject-predicate-object triples, it preserves full syntactic structure as ordered, recursive hyperedges. The README’s example shows “Einstein first published the theory of relativity in 1905” rendered as nested blocks rather than a sparse graph.

The interesting bit

The project treats hyperedges as first-class citizens—recursive, ordered, and hyperlink-like. This is pitched as foundational work, funded by CNRS and an ERC grant, suggesting academic rigor behind the abstraction. The explicit Python choice signals intent to plug into NumPy-era scientific computing stacks rather than chase performance benchmarks.

Key highlights

• Represents natural language as recursive, ordered hyperedges rather than flat triples
• Python-native, targeting integration with existing scientific libraries
• MIT licensed; funded by CNRS and ERC Consolidator Grant #772743 (Socsemics)
• Documentation and manual hosted externally at hyperquest.ai/hyperbase
• Accepts pull requests; asks for tests with changes

Caveats

• The README is essentially a landing page; actual API surface, performance characteristics, and maturity level are unclear without visiting the external manual
• No code examples, version info, or dependency list in the repository itself
• One contributor-friendly image (the Einstein hyperedge diagram) and one funder logo; the project appears to be early-stage infrastructure

Verdict

Worth a look if you’re building knowledge representation systems and have hit the expressiveness ceiling of RDF-style triples. Skip it if you need battle-tested tooling or can’t stomach visiting an external site for basic documentation.