Automatisk extrahering av ordleder

Abstract

A typical trait of Swedish word formation are words which are productive in word formation. They are not bound morphemes because they are either root morphemes or composites·. They are not lexemes because they are used in a similar way as bound morphemes. These bound lexemes, as they are called here, constitute a crucial part of a compounding dictionary, particularly of a bilingual one. Potential bound lexemes are traced and automatically selected from a large text corpus of Swedish. Two parameters are established to make the extraction precise: compounding ability and compounding productivity. The first one provides measure of the relation between occurrences of a word as compound component and as an independent word. The second one is based on how many collocates it cooccurs with in compounds.