Abstract Malouf
In this talk, we will discuss on-going work in the development and evaluation of statistical disambiguation models for unification-based parsing of Dutch within the framework of the Alpino project. The Alpino grammar (Bouma, et al. 2001) is a broad-coverage Head-Driven Phrase Structure Grammar of Dutch which we have augmented with a maximum-entropy probability model to allow selection of the most likely parse (Abney 1997, Johnson, et al 1999). We will present results of our current work on combining supervised and unsupervised training and on applying resampling techniques to improve feature selection and parameter estimation.Last modified $Date: 2001/10/04 13:39:47 $ by Parlevink Webmaster