UPOS

UPOS, short for Universal Part-of-Speech tags, is a coarse-grained tagset used in the Universal Dependencies (UD) framework. It provides a language-agnostic set of POS categories to annotate tokens across languages, enabling cross-linguistic comparisons, alignment of parallel texts, and standardized evaluation in natural language processing and linguistic annotation.

The UPOS tagset comprises 17 tags: ADJ, ADP, ADV, AUX, CCONJ, DET, INTJ, NOUN, NUM, PART, PRON, PROPN, PUNCT, SCONJ, SYM, VERB, and X. They represent universal grammatical categories that map to language-specific annotations during UD annotation. Morphological details are captured separately in features (FEATS); UPOS provides a stable core tag for cross-language analysis.

In practice, UPOS is applied to annotated corpora known as UD treebanks, which are used to train POS taggers and dependency parsers. The tagset supports cross-language research and multilingual NLP by providing a consistent standard across languages. Example: The sentence "The cat sat on the mat." would be tagged as The/DET cat/NOUN sat/VERB on/ADP the/DET mat/NOUN ./PUNCT.