kenon

Semantic and co-occurrence graphs for midsized texts.

Kenon builds weighted graphs from text using corpus-internal statistics only. No neural models or external training data required.

Core concepts

• Co-occurrence graphs: Tokens as nodes, skip-gram window co-occurrence as edges
• Semantic graphs: Tokens as nodes, cosine similarity from embeddings as edges
• Backbone extraction: Disparity filter to keep only statistically significant edges

Quick links

• Quickstart — clean install to backbone graph
• Tutorial: word-association graphs — find paths between concepts
• Examples — full, runnable scripts
• Troubleshooting — common errors and fixes
• API Reference — every public function and its contract

---

Made by Crow Intelligence