Select a lesson to begin
Choose a lesson from the list on the left, or click "Custom text" to paste your own.
Entity types to extract
?
GLiNER — Generalist and Lightweight Model for Named Entity Recognition.
A fast, self-hosted zero-shot model. You give it a list of entity type labels and it finds matching spans in the text — no retraining needed.
⚡ ~180ms per document
✓ Returns confidence scores (the % shown on each entity)
✓ Runs fully locally — no API calls
✓ Recommended for production ingest
A fast, self-hosted zero-shot model. You give it a list of entity type labels and it finds matching spans in the text — no retraining needed.
⚡ ~180ms per document
✓ Returns confidence scores (the % shown on each entity)
✓ Runs fully locally — no API calls
✓ Recommended for production ingest
?
Gemma3:4b — a local large language model running via Ollama.
Reads the full text and uses language understanding to extract entities. More flexible but significantly slower.
🐢 ~2500ms per document (14× slower than GLiNER)
✗ No confidence scores — extracts or doesn't
✓ Runs fully locally — no API calls
✓ Included here as a comparison baseline
Reads the full text and uses language understanding to extract entities. More flexible but significantly slower.
🐢 ~2500ms per document (14× slower than GLiNER)
✗ No confidence scores — extracts or doesn't
✓ Runs fully locally — no API calls
✓ Included here as a comparison baseline
Extracted Entities
Run an extractor to see results.