Halvautomatisk udvælgelse af lemmakandidater til en nyordsordbog
Jakob Halskov · LexicoNordica · 2010
One of the key tasks of the Danish Language Council (Dansk Sprognævn)is to monitor, record and document linguistic changes inthe Danish language. To assist in this task, a prototype neologismdetector called the Ordtrawler (Word trawler) is being developed.The system processes large amounts of text and extracts candidateneologisms using a combination of simple filters, collocational statisticsand so-called neology markers. A thorough evaluation of theOrdtrawler indicates that neology markers are good for optimizingsystem precision, but at the expense of recall. Certain types of noisesuch as technical terms and semantically transparent compoundsremain to be tackled, but diachronic frequency profiling may help.