Clustering verbs semantically according to their alternation behaviour
Sabine Schulte im Walde · 2000
Verbs were clustered semantically on the basis of their alternation behaviour, as characterised by their syntactic subcategorisation frames extracted from maximum probability parses of a robust statistical parser, and completed by assigning WordNet classes as selectional preferences to the frame arguments. The clustering was achieved (a) iteratively by measuring the relative entropy between the verbs' probability distributions over the frame types, and (b) by utilising a latent class analysis based on the joint frequencies of verbs and frame types.