Semantisets

Semantisets are a theoretical construct in formal semantics and computational linguistics that model meaning as collections of representations rather than single predicates. A semanset is a subset of a semantic universe S, where S comprises basic meaning elements such as features, propositions, or context-dependent representations. Each lexical item w is associated with one or more semantic contents M(w) ⊆ S, representing its possible meanings in a given language. A semanset groups together those representations that share a given semantic property or role.

In practice, semantisets support a modular view of meaning. For example, the word bank can be associated with two broad semantisets: one containing meanings related to a financial institution and another containing meanings related to a riverbank. More refined semantisets can encode features like animacy, agentivity, spatial relation, or temporal scope, enabling finer-grained interpretation control.

Semantics composition is expressed through set operations on semantisets. Union combines alternative senses; intersection captures senses compatible with contextual constraints; difference excludes senses ruled out by discourse or world knowledge. During sentence interpretation, context-derived semantisets from syntax, pragmatics, and background knowledge intersect to yield a target semanset for potential meanings. Disambiguation then selects the most plausible member of that target set.

Applications of semantisets include lexical disambiguation, semantic parsing, knowledge representation, and cross-lingual alignment in multilingual corpora. They offer a discrete, interpretable complement to continuous vector embeddings, providing a transparent framework for tracking which semantic features are active in a given interpretation and how meaning changes with context. The concept is used in theoretical discussions and exploratory computational models to formalize semantic composition and interpretation.