| In this talk I present a probabilistic generative model for joint
morphological and syntactic parsing. The formal model is defined in
language independent terms and is based on typological principles,
while the statistical learning component gathers language specific
information from linguistically annotated corpora. We present parsing
experiments using the acquired statistical models on three, typologically
different, languages, and show that the parsing results are on a par with, or better
than, language-specific or task-specific implementations. |