Using chunked corpora for the acquisition of collocations and idiomatic expressions

This paper1 discusses the use of recursive chunking of large German corpora (over 300 million words) for the identification and partial classification of significant lexical cooccurrences of adjectives and verbs. The goal is to provide a fine-grained syntactic classification of the data at the levels of subcategorization and scrambling. We analyze the combinatory preferences of adjectives with verbs in predicative(-like) constructions subcategorizing a finite or infinite clause. Adjective+verb collocations have so far not been analyzed in much detail. The corpus data suggest that here is in many cases a clear correlation not only between collocational preferences, distributional properties, and subcategorization frames, but also between all of them and semantic classes of adjectives. Some adjective+verb combinations seem to be idiomatic, or lexicalized as complex predicates. The availability of a robust broad coverage chunker for German, along with an automatic partial classification into different phenomena is a prerequisite for this kind of detailed analysis of collocations and idiomatic expressions. 1This work has been carried out in the framework of the Transferbereich 32: Automatische Exzerption, a DFG-funded project aiming at the creation of support tools for the corpus-based updating of printed dictionaries in lexicography carried out in cooperation with the publisher Duden BIFAB AG and Langenscheidt KG.