Identify People, Organizations, Locations, and other named entities from any text. Runs locally in your browser using BERT-base-NER.
100% Private. Your text is never uploaded. The model runs locally via WebAssembly.
Runs entirely in your browser. Your data never leaves your device.
We benchmarked BERT-base-NER on a curated set of news sentences:
| Metric | Value |
|---|---|
| Model size (INT8) | ~108MB |
| F1-score (CoNLL-2003) | 89.1% |
| Avg inference time | ~200ms / sentence |
| Max sequence length | 512 tokens |
| First load (cold) | ~7s |
Last verified: September 2026. MacBook Air M2.
The BERT-base-NER model extracts 4 entity types: PER (Person — names of people), ORG (Organization — companies, institutions), LOC (Location — cities, countries, landmarks), and MISC (Miscellaneous — other proper nouns like titles, events, products).
No. The NER model runs entirely in your browser using WebAssembly. Your text never leaves your device. The ~108MB model downloads once from CDN and caches locally in IndexedDB.
BERT-base-NER achieves ~89% F1-score on the CoNLL-2003 benchmark for 4-class NER. Performance may vary for short text, informal writing, or non-English content.
Yes. Copy entities as JSON (with text, type, confidence, start/end positions) or CSV for spreadsheet analysis.