Exploiting non-linear probabilistic models in natural language parsing and reranking

Diplôme Thèse Exploiting non-linear probabilistic models in natural language parsing and reranking eng 2008-01-28 author Titov Ivan 135628 director Pellegrini Christian 24713 director Henderson James 529988 10.13097/archive-ouverte/unige:8495 urn:nbn:ch:unige-84958 Sc. 3943 The thesis considers non-linear probabilistic models for natural language parsing, and it primarily focuses on the class of models which do not impose strict constraints on the structure of statistical dependencies. The main contribution is the demonstration that such models are appropriate for natural language parsing tasks and provide advantages over the use of standard 'linear' methods. We demonstrate this, first, by showing that though exact inference is intractable for the studied class of models, there exist accurate and tractable approximations. Second, we show that using non-linear representations results in powerful feature induction methods simplifying construction of parsers for new problems and domains, and leading to the state-of-the-art performance. Also, we demonstrate that the latent space induced by the model can be exploited in discriminative rerankers, and that this results in the significant improvement both in the standard parsing settings and in domain adaptation. Natural language parsing Syntax Dependency parsing Constituent parsing Natural language learning Bayes risk minimization Dynamic Bayesian networks Variational approximations Dewey 025.063 13 Centre universitaire d'informatique 0D3 32 Département d'informatique 18 false fe56a113-e68c-4cc9-ab45-58328a27866d Thesis Public thesis.pdf 1574202 application/pdf openAccess Centres et instituts / Centre universitaire d'informatique Faculté des sciences / Département d'informatique

Title	Published in	Year	Views	Downloads

Exploiting non-linear probabilistic models in natural language parsing and reranking		2008	813	406
Large Margin Multiple Hyperplane Classification for Content-based Multimedia Retrieval	Proceedings of the 1st Workshop on Machine Learning Techniques for Processing Multimedia Content, MLMM'05	2005	365	0

Archive ouverte UNIGE

Titov, Ivan