Digital Lexicon Hub

A natural-language-processing task that identifies and classifies named entities such as people, organizations, locations and dates in text.

NER turns unstructured text such as emails, contracts, reviews and support tickets into structured fields, which supports routing, search, compliance checks and analytics without manual tagging.

A legal operations team runs NER over 5,000 vendor contracts to extract party names, dates and amounts into a spreadsheet, then has reviewers check a random sample of 200 extractions before trusting the output.

Models trained on general news text often miss domain entities such as product codes or drug names. Ambiguous names, nested entities and new terms cause errors, and NER identifies mentions without linking them to a specific real-world record.

NER is typically treated as sequence labeling: each token receives a tag using a scheme such as BIO or BILUO that marks the beginning, inside and end of an entity span and its type. Models range from conditional random fields to transformer encoders fine-tuned on annotated corpora. Performance is measured with precision, recall and F1 at the entity level, and results can be passed to entity linking to match mentions to a knowledge base.