What do we get from extracting collocations?
Daria Kormacheva · Journal of Research Design and Statistics in Linguistics and Communication Science · 2015
This paper applies linguistic analysis to the results from the automatic extraction of multiword expressions in order to understand whether they are reliable from the theoretical point of view. The nature of the extracted units is discussed and illustrated with examples of Russian prepositions: first classified according to I. Mel’?uk’s theory (1995) and then re-analysed using the notion of constructions. The corpus-driven approach reveals the shortcomings in the prevalent way of describing multiword expressions in terms of strict classes, and the present paper can be thought of as providing a theoretical basis for the development of a new approach to their description.