PROPN

PROPN is a part of speech tag used in linguistic annotation schemes to label proper nouns—names that refer to unique entities such as people, organizations, places, or titled items. In Universal Dependencies and related corpora, PROPN marks tokens that denote specific entities rather than general classes of nouns. This contrasts with NOUN, which covers common or countable nouns.

In practice, PROPN can cover single tokens like Paris or Tesla, as well as multiword names such as New York City or the United Nations. Capitalization in the source text is a strong cue but not reliable, and annotation schemes often decide how to represent multiword proper nouns: as a sequence of PROPN tokens with compound relations, or as a single multiword token when supported by the tokenizer.

PROPN is used in parsing and downstream tasks such as named entity recognition and information extraction, helping systems distinguish named entities from ordinary nouns. However, PROPN by itself does not assign entity type; separate labels (PERSON, LOC, ORG, GPE, etc.) or NER outputs may provide that information.

Cross-linguistically, the set of items tagged PROPN varies by language and script. Some older English resources use the tag NNP (proper noun) from the Penn Treebank instead of PROPN, but modern UD conventions prefer PROPN. Proper noun tagging interacts with morphology in inflected languages, where case, number, or gender features may accompany the token.