Orð og tunga - 01.06.2012, Page 14

4
Orð og tunga
The WordNet database is therefore structured in terms of a number
of sense relations which appear to be psychologically relevant in the
characterisation of word meaning. Further the database is organised
around part of speech, on the basis of evidence that word storage
in the mental lexicon is sensitive to part of speech. The current dis-
cussion relates to WorciNet 3.0, which contains around 155,000 word
forms (unique strings), of which just over 115,000 are nouns; the rest
are verbs, adjectives and adverbs. In the following sections, we will
review some of the main lexical sense relations that determine the
organisation of WordNet.
2.2 Synonyms and synsets
The basic building block of WordNet is the synset or "set of syno-
nyms" (Icelandic: samheiti; Greek: syn 'same' + onyma 'name'). In
WordNet, synonymy is defined as having the same sense in a particu-
lar context.
(1) the nurse gave him a flu shot/injection/*pellet
• synset: = (shot, injectionj
(2) the shot/pellet/*injection buzzed past his ear
• synset: = (shot, pelletj
Sentence (1) identifies a particular "sense", glossed in WordNet 3.0 as
"the act of putting a liquid into the body by means of a syringe". This
sense can be expressed by shot and by injection but not by pellet; shot
and injection are therefore synonyms and form a synset. Sentence (2)
identifíes another "sense", glossed in WordNet 3.0 as "a solid missile
discharged from a fírearm". This second sense can be expressed by
sliot and by pellet but not by injection. This illustrates two important
points about the organisation of WordNet.
First, the basic building block of the network is in fact a particular
sense or concept; that sense can be expressed by one or more different
word forms. This is very different from a traditional dictionary, whose
basic building block is the word itself: the forms shot and injection
would be listed separately in a traditional dictionary and each would
be listed with the relevant sense as part of its entry. In WordNet, the
sense itself represents a unique entry and the forms associated with