Extraction and recognition of Polish multiword expressions using Wikipedia and finite-state automata

Abstract

Linguistic resources for Polish are often missing multiword expressions (MWEs) - idioms, compound nouns and other expressions which have their own distinct meaning as a whole. This paper describes an effort to extract and recognize nominal MWEs in Polish text using Wikipedia, inflection dictionaries and finite-state automata. Wikipedia is used as a lexicon of MWEs and as a corpus annotated with links to articles. Incoming links for each article are used to determine the inflection pattern of the headword - this approach helps eliminate invalid inflected forms. The goal is to recognize known MWEs as well as to find more expressions sharing similar grammatical structure and occurring in similar context.