Processering af en synonym-komponent i en flersproget ordbog

Abstract

The ISLEX dictionary combines Icelandic as a source language (SL) and six Nordic languages as target languages (TLs) in a single database. This article describes the process of how Icelandic synonyms can be extracted and grouped together by using the TLs’ equivalents. Each word class is processed independently. Selected material, a list of lemmas with their ID number plus their TLs, is exported from the database, going through the languages one at a time. This produces long lists of results which consist of the Icelandic headwords and their equivalents. The lists are then processed in certain ways. In cases where a lemma has more than one sense there is a danger that the boundaries between the senses become blurred so that the words in those groups tend to get mixed up. An example is the Icelandic noun verslun which means both ’shop’ and 'trade’, and is therefore divided into two senses in the dictionary. All the material comes from within the dictionary and no external data is used. It means that all the synonyms obtained in this way are necessarily also lemmas. Despite this we consider the method justifiable because it is quick and simple, and in many instances it gives very good results, i.e. groups of synonyms with varying degrees of exactness.