Lexical & Compositional Semantics

← Back to Domain-Specific Families

Abstractions about how word meaning is structured, composed, and represented, including lexical relations and processes such as hypernymy, back-formation and lexicalization, formal and distributional semantic models like componential analysis, latent semantic analysis and word embeddings, and applied semantic tasks such as word-sense disambiguation and semantic parsing.

24 abstractions in this family — domain-specific abstractions that sit near one another in structural-signature space (k-means over structural-signature embeddings). Each is shown with its short description.

• Alternative semantics — A formal semantics in which expressions denote sets of alternatives rather than single values, enabling compositional accounts of focus, questions and indefinites.
• Autological word — A word or expression that itself possesses the property it denotes.
• Back-formation — The creation of a new lexeme by removing or replacing a segment interpreted as an affix from an existing word.
• Bag-of-words model — A text representation that records which tokens occur and often their counts or weights while discarding word order and most grammatical structure.
• Collocational restriction — A lexical constraint in which a word selects or acquires a specialized sense only in combination with particular neighboring words.
• Componential analysis — A structural-semantic method representing word meanings through contrasts over a set of binary or multivalued semantic features.
• Concept search — Information retrieval that ranks unstructured documents by semantic similarity to the ideas expressed in a query rather than exact word overlap alone.
• Cryptotype — A covert grammatical or semantic category whose effects appear in distribution and combination even though it has no overt morphological marker.
• Denotation — The entity, value, extension, truth condition, or semantic object that an expression is assigned under an interpretation, distinguished from associated connotations and many pragmatic effects.
• Hypernymy and hyponymy — The inverse lexical-semantic relations linking a more general term to a more specific term whose meaning is included beneath it.
• Latent semantic analysis — A distributional-semantic technique that applies truncated singular-value decomposition to a term–document matrix so terms and documents are represented in a shared lower-dimensional latent space.
• Lexicalization — The process by which a meaning, phrase or word-formation pattern becomes established as a conventional lexical item or stored unit of a language.
• Lexicostatistics — A comparative-linguistic method that quantifies shared basic-vocabulary cognates among languages to estimate degrees of lexical relationship.
• Meaning–text theory — A formal linguistic framework modeling language as a multilevel correspondence between semantic representations and surface text through intermediate syntactic and morphological structures.
• Plural of abstraction — A grammatical use of plural morphology on a concrete noun or manifestation to denote an abstract quality, state or totality of its instances.
• Quantificational variability effect — The interpretive correspondence between adverbial quantification over situations or events and determiner quantification over participating individuals.
• Relationship extraction — An information-extraction task that detects entity pairs or other arguments in artifacts and assigns the semantic relation expressed between them.
• Semantic compression — Lossy reduction of lexical diversity by replacing expressions with a smaller concept vocabulary while attempting to preserve their meaning.
• Semantic gap — The mismatch in meaning between descriptions formed in different representation systems, especially between human contextual concepts and the formal structures available to a computer system.
• Semantic parsing — The computational task of mapping a natural-language utterance into a machine-interpretable formal meaning representation that supports execution, inference or structured comparison.
• Skunked term — A word or phrase whose meaning is in unstable transition or controversy, making ordinary use likely to distract, divide, or mislead readers regardless of which competing sense the writer intends.
• Specified subject condition — A historical transformational-grammar constraint blocking certain rule applications across a clause or phrase containing an overt lexical subject distinct from the moved or related element.
• Word embedding — A learned vector representation of words or tokens in which geometric relations encode patterns of distributional or task-specific similarity.
• Word-sense disambiguation — Select the contextually intended sense of an ambiguous word from a declared sense inventory, while keeping inventory granularity and annotation uncertainty separate from model accuracy.