Synset

A synset (short for synonym set) is a group of one or more synonyms that share a common meaning in some context. Synsets are a fundamental concept in computational linguistics and lexical semantics, most notably used in the WordNet lexical database of English.^[1]

Overview

The concept of synsets was introduced by George A. Miller and colleagues as part of the development of WordNet at Princeton University in the mid 1980s.^[1] Each synset represents a distinct lexicalized concept and consists of one or more lemmas that are synonymous in at least one sense. For example, the synset {car, automobile, motorcar, auto} represents the concept of a self-propelled vehicle with four wheels, while the synset {car, railcar, railway car, railroad car}, also including the lemma car, represents the concept of a wheeled vehicle adapted to the rails of railroad.

Synsets are organized into a network via semantic relations such as hypernymy (e.g. {car, automobile, motorcar, auto} is a kind of {motor vehicle, automotive vehicle}), meronymy (e.g. {car, automobile, motorcar, auto} has part {car door}), and entailment (e.g. {snore, saw wood, saw logs} entails {sleep, kip, slumber, log Z's}). These relations allow computational models to capture structured knowledge about word meaning.

While synsets originated in the English WordNet, the concept has been extended to other languages and multilingual projects such as EuroWordNet,^[2] BabelNet,^[3] and the Global WordNet initiative.^[4]

Applications

Synsets have been shown to play an important role in some areas of Natural Language Processing, providing a bridge between lexical items and conceptual meaning, enabling:^[5]

• Word sense disambiguation
• Information retrieval
• Text mining
• Machine translation
• Ontology construction for the semantic web

Metadata

In metadata, a synonym ring or synset, is a group of data elements that are considered semantically equivalent for the purposes of information retrieval.^{[citation needed]} These data elements are frequently found in different metadata registries. Although a group of terms can be considered equivalent, metadata registries store the synonyms at a central location called the preferred data element.

Example

The following are considered semantically equivalent and form a synonym ring:

foaf:person
gjxdm:Person
niem:Person
sumo:Human
cyc:Person
umbel:Person

Note that each data element has two components:

1. Namespace prefix, which is a shorthand for the name of the metadata registry
2. Data element name, which is the name of the object in each of the distinct metadata registry

Expressing a synonym ring

A synonym ring can be expressed by a series of statements in the Web Ontology Language (OWL) using the classEquivalence or the propertyEquivalence or instance equivalence statement – the sameAs property.

See also

• WordNet
• Lexical semantics
• Ontology (information science)
• Word sense disambiguation
• Data Reference Model
• Metadata
• Vocabulary-based transformation