Extracting Idiomatic Hungarian Verb Frames

We describe a machine learning method for collecting idiomatic fixed stem verb frames. Firstly we collect frequent frame candidates from the output of a partial parser, secondly we apply a certain idiomaticity metric to the list to get the most idiomatic frames. Running our implemented system we get a list of ten thousand frames of more than 900 verbs which will be translated to English and used as a resource in a Hungarian-to-English machine translation system.