Lexemer

Lexemer is a theoretical unit in lexical semantics and computational linguistics designed to capture a single sense of a word together with its grammatical and semantic properties. It is intended as the smallest context-ready unit that can be linked to a specific sense in a lexical database or NLP model. The concept distinguishes between a word’s overall lexical item (lexeme) and the individual senses it encodes; each sense is represented by a separate lexemer.

A lexemer typically includes several components: the lemma or base form, a sense_id that uniquely identifies the specific meaning, part_of_speech, semantic_features, syntactic_frames that specify allowable argument structures, usage constraints, and optional frequency or confidence measures. These elements enable precise, sense-specific representations that support disambiguation, parsing, and downstream processing tasks.

Relation to other notions: a lexeme denotes the complete set of all forms and senses of a word, whereas a lexemer corresponds to one sense within that set. Sememes are a broader category of minimal semantic units, and lexemers can be viewed as sense-specific, syntacto-semantically enriched sememe instances. In practice, lexemers are used in lexical databases, word sense disambiguation, machine translation, and semantic parsing.

Examples: the word bank can have at least two lexemers: bank financial_institution (sense_id FIN-01) with syntactic frames like a bank [det] [to be] funded, and bank riverbank (sense_id RIV-02) with frames like the bank [of] a river. These distinct lexemers reflect different meanings tied to the same surface form.

Applications and challenges: lexemers enable fine-grained sense representation and disambiguation but defining sense boundaries and obtaining robust annotations remain challenging. Careful alignment with annotation schemes and corpora is required. See also Lexeme, Sememe, Word sense disambiguation, Lexical database.