Halvautomatisk udvælgelse af lemmakandidater til en nyordsordbog

Resumé

One of the key tasks of the Danish Language Council (Dansk Sprognævn)
is to monitor, record and document linguistic changes in
the Danish language. To assist in this task, a prototype neologism
detector called the Ordtrawler (Word trawler) is being developed.
The system processes large amounts of text and extracts candidate
neologisms using a combination of simple filters, collocational statistics
and so-called neology markers. A thorough evaluation of the
Ordtrawler indicates that neology markers are good for optimizing
system precision, but at the expense of recall. Certain types of noise
such as technical terms and semantically transparent compounds
remain to be tackled, but diachronic frequency profiling may help.