In this paper, a complex-feature- and MDL-based model for acquisition of VN-construction structure templates is put forward. First, a verb classification tree is created using statistical decision tree model. Then, the tree is pruned based on MDL (minimum description length) principle. Finally, structure templates are derived based on the verb classification tree. The experiments show that using the structure templates acquired with the model to recognizing VN-structure, the system has its advantages over the model based on the sense and the MLE (maximum likelihood estimation) principle in precision and recall.