Named Entity Extractor
100% private — runs on your device, never uploaded. Works offline once loaded.
Paste text and highlight the people, organizations and locations in it using an on-device named-entity recognition model. Your text never leaves the browser.
On-device named-entity recognition
This tool runs a BERT model fine-tuned for named-entity recognition (NER) directly in your browser. It scans your text and tags spans as people (PER), organizations (ORG), locations (LOC) or miscellaneous (MISC), then highlights them and groups the unique entities by type.
Everything runs on your device, so contracts, emails and research notes you analyze are never uploaded — useful when the document contains names and details you need to keep private.
How the tagging works
NER models label word pieces with begin/inside tags, which this tool merges back into whole entities using the model’s character offsets. That is why “New York” is returned as one location rather than two tokens, and multi-word names stay intact.
Common uses and caveats
NER is handy for redaction checklists, building contact lists from documents, or summarizing who and what a text is about. It is not perfect: unusual names, creative capitalization, or domain-specific terms can be missed or mislabeled, so verify results when accuracy matters.
- Spot names and places to redact before sharing a document
- Pull the organizations mentioned across a report
- Get a quick who/where overview of an article
Frequently asked questions
Is my text uploaded?
No — the model runs entirely in your browser and your text is never sent anywhere.
Which entity types are recognized?
Persons, organizations, locations and a miscellaneous category for other proper nouns.
Why are some names missed?
The model is compact and English-trained, so rare names, unusual casing or niche terms can be missed — review important results.
Does it keep multi-word names together?
Yes. The tool merges the model’s sub-word tags into whole entity spans using character offsets.
Why is the first run slow?
A one-time model (~110 MB) downloads and caches on first use.
Does it work offline?
Yes, once the model is cached it runs without a network connection.
Advertisement