Automatic Recognition of Multi-Word Expressions in Dutch

Abstract

In this bachelor's thesis I research the possibility of finding multi-word expressions (MWEs) in a large Dutch corpus called LASSY using the statistical distance methods called Point Mutual Information and Salience. I compare the achieved results with results from an already existing toolkit called the MWEtoolkit. The conclusion is that MWEs can not be detected with the chosen methods in the chosen way, but that the method might be useful when combined with other detection methods.