Fra ordbog til wordnet. Hvordan udmøntes en traditionel ordbogsdefinition i en formaliseret wordnetbeskrivelse?

Abstract

In this article we discuss the problems involved in transforming traditional dictionary definitions into formalised descriptions meant for computational use. The starting point is the making of the Danish wordnet, DanNet, which is based on a large monolingual dictionary. In the process of extracting semantic information from the entries it turns out that much information is already present in the dictionary text, but a lot of useful information is only implicitly there or presupposed. The article falls into three parts: firstly, the principles of lexicographic definitions are discussed, in particular the ones used in The Danish Dictionary (Den Danske Ordbog). Secondly, it is demonstrated how definitions are “translated” into semantic relations in the wordnet, and examples of how missing information is added are given. Finally, the idea of making general dictionaries and wordnets in one operation is put forward and discussed.