Leksikosemantiliste suhete hägusus Eesti Wordnetis
Fuzzy lexical-semantic relations in Estonian Wordnet
Author(s): Heili Orav, Sirli Zupping, Kadri VareSubject(s): Lexis, Semantics, Computational linguistics, Finno-Ugrian studies
Published by: Teaduste Akadeemia Kirjastus
Keywords: lexical semantics; computational lexicology; language resources; computational linguistics; Estonian;
Summary/Abstract: This paper gives an overview of the principles of wordnets in general and focuses mainly on the Estonian Wordnet (EstWN). The latest version of EstWN consists of more than 72,000 concepts and 51 different lexical relations are used to form a network of more than 230,000 semantic relations between concepts. The main relations that EstWN uses are hyperonymy, meronymy, involvement and fuzzynyms (in Princeton WordNet, for example, hyperonymy is the most implemented relation). Of course the richness of different types of relations creates problems and unclear determination of these relations. In case of hyperonyms the developers of EstWN have encountered problems in choosing preferably only one suitable hyperonym for each concept. When dealing with meronymy the more specific relations – involved location, involved direction (both source and target direction) – are inconsistently determined. There are, however, no significant problems with involved instrument and involved agent relations. In PWN there is no involved location of direction relation explicitly available. Meronymy relations are often associated with the problems of connecting encyclopedic concepts to those of general language, for example how to connect the concept ‘bird’ to a specific bird species. In EstWN the general language vocabulary is well covered, specific domain vocabularies are also incorporated (architecture, medicine, economy etc.) and it would be useful to connect specific vocabulary to general language vocabulary. The paper proposes that the answer to this problem could be the complementary information provided from domain labels. The last semantic relation discussed in this paper deals with fuzzynymy, since this is the third used relation in EstWN. Fuzzynymy is a free association relation, but it is clear that some groups form out of the fuzzynymy relation that can be defined as new specific relations in Estonian. Recently EstWN has become an increasingly used resource in Estonian language technology, and as such it is important to improve the quality and consistency of relations in addition to increasing the amount of concepts in EstWN in different domains.
Journal: Emakeele Seltsi aastaraamat
- Issue Year: 2014
- Issue No: 60
- Page Range: 171-194
- Page Count: 24
- Language: Estonian