Back to Main Conference 2004
LREC 2004main
Semi-Automatic Construction of a Question Treebank
Proceedings of the Fourth International Conference on Language Resources and Evaluation (LREC 2004)
Abstract
A method for the semi-automatic construction of a question treebank is presented. We exploit linguistic knowledge like grammatical functions, constituent structure and the relatively strict word order of English encoded in the Penn Treebank to generate semi-automatically questions. The outcome is a treebank of questions which might be useful for developing better tagging and parsing models for processing questions. We show that it is feasible to reuse the Penn Treebank. At the current stage our treebank comprises about 7000 questions, which can be easily extended.