Lemmat

A lemmat is a basic unit in some linguistic and computational contexts that combines a word's lemma with its morphosyntactic information. A lemma is the canonical dictionary form of a lexeme; a lemmat thus represents the pair consisting of that base form and its grammatical features, such as part of speech, tense, number, or case. In annotated corpora and lexicographic databases, lemmats are used to represent the underlying meaning and grammatical behavior of surface forms, enabling consistent analysis across inflected or derived variants.

The term is not widely standardized and is mostly encountered in discussions of corpus annotation schemas, lemmatization pipelines, and lexicographic data models. It is distinct from a plain lemma (which stores only the base form) and from a token (the exact surface form as it appears in text). A lemmat can be seen as a compact representation used by NLP systems to link inflected tokens to their canonical base form with features.

Examples: The surface form "running" maps to the lemmat ("run", "VBG"); "better" maps to ("good", "JJR"). In practice, lemmats can support search, disambiguation, and linguistic analysis by preserving both base form and grammatical information.

See also: Lemma (linguistics); Lemmatization; Morphology; Part of speech tagging.