Browse by topic
Type of publication
Confidence Measures for Stochastic Parsing. Proceedings of the International Conference Recent Advances in Natural Language Processing, 2007. pp. 58-63.In this work the use of confidence measures for detecting errors of stochastic parsing is explored. The confidence measures are based on posterior probabilities computed over a list of the n-best parse trees. Several confidence measures are proposed and a naive Bayes model is also considered. The proposed confidence measures are tested with the Charniak parser and the Penn Treebank corpus.