The Design of Wordnet

network of diseases
Image by DrJohnBullas via Flickr

Introduction

Wordnet is based on a model of the organization of human memory called a  Semantic Network (see  Collins and Quillian). Such networks represent semantic relations among concepts as in the diagram on the right. We have knowledge of tens of thousands of concepts which we can usually recall effortlessly.

As a lexical database of English, Wordnet is concerned with nouns, verbs, adjectives and adverbs. Unlike a conventional dictionary, words are grouped together into sets of synomyns  called synsets. A synset is Wordnet’s basic building block and contains one or more words. Some possible synsets are:

  • {shape, form}
  • {gazing, shaving, skimming}
  • {telegraphically, tersely}
  • {takeoff}

Within the context of Semantic Nets, a Synset represents a concept and each word within the set has a sense which defines that concept. For instance, the concept of the spatial arrangement of something as distinct from its substance can be referred to as form or shape.

So what are the relationships among words that Wordnet defines? The designers of Wordnet decided to base the relationships between Synsets on hierarchical lexical and semantic relationships most of which end in onym.  The most common semantic relations defined are:

Synonym: a word which may replace another word in a sentence without changing the meaning. In Wordnet, a synset (synonym set) contains words that are equivalent in meaning or nearly so.

Antonym: a word with the exact opposite meaning of another word.

Hypernym: a word which defines the class to which other equally-ranked items belong, such as the generic word “tree” for “oak” or “hazel.” For verbs we have walk which is a way of moving on foot and a generic term for verbs like creep, totter and amble.

Hyponym: one of the equally-ranked members of a generic class or group of items., for example “lily” or “violet” in the class of “flowers.”

Holonym: a word for the whole of which other words are a part. For instance house contains roof and window and is therefore the whole in this whole/part relationship.

Meronym: a word that names a part that belongs to a whole or larger entity. All of these are part-whole relationships:

  • “door” or “window” in “house”
  • “engine” or “steering-wheel” in “car”

Troponym: a verb that indicates more precisely the manner of doing something by replacing a verb of a more generalized meaning, thus the verb to stroll indicates a more leisurely, casual manner of to walk.

Entailment: a relationship between two words where the truth of one word requires the truth of the other. For instance snoring requires sleeping to be true – the verb to sleep is entailed by to snore.

Relationships for nouns

  • synonym/antonym
  • hypernym/hyponym
  • holonym/meronym

Relationships for verbs

  • synonym/antonym
  • hypernym
  • troponym
  • entailment

Relationships for adjective

  • synonym/antonym

Relationships for adverbs

  • synonym

WordNet was developed by the Cognitive Science Laboratory at Princeton University, under the direction of George A. Miller (Emeritus).

Related articles by Zemanta

Leave a comment