Title

Functional arguments detection using decision trees

Abstract

This paper shows how decision trees can help addressing the problem of functional arguments detection, namely logical subjects. On a set of 14 target-verbs widely used in verb studies, verb-centered models provide almost similar results with all-verbs models. When traces in training data are solved the size of training increases with about 10% while the performance remains the same with the advantage of a more reliable estimation.

Publication Title

Proceedings of the International Conference on Artificial Intelligence, IC-AI'04

This document is currently not available here.

Share

COinS