Priberam Machine Learning Lunch Seminar: "Posterior Regularization Framework: Learning Tractable Models with Intractable Constraints"

Speaker: João Graça (L2F, INESC-ID)
Venue: IST Alameda, Sala PA2 (Edifício de Pós-Graduação)
Date: Tuesday, June 22th, 2010
Time: 13:00
Lunch will be provided

Abstract:
Unsupervised Learning of probabilistic structured models presents a fundamental trade- off between richness of captured constraints and correlations versus efficiency and tractability of inference. In this thesis, we propose a new learning framework called Posterior Regulariza- tion that incorporates side-information into unsupervised estimation in the form of constraints on the model’s posteriors. The underlying model remains unchanged, but the learning method changes. During learning, our method is similar to the EM algorithm, but we solve a problem similar to Maximum Entropy inside the E-Step to enforce the constraints. We apply the PR framework to two different large scale tasks: Statistical Word Alignments and Unsupervised Part of Speech Induction. In the former, we incorporate two constraints: bijectivity and symme- try. Training using these constraints produces a significant boost in performance as measured by both precision and recall against manually annotated alignments for six language pairs. In the latter we enforce sparsity on the word tag distribution which is overestimated using the default training method. Experiments on six languages achieve dramatic improvements over state-of-the-art results.

Bio:
I am currently a 4th year PhD student (with MSc degree) in Computer Science Engineering at Instituto Superior Técnico, Technical University of Lisbon and a visiting student at University of Pennsylvania. My advisors are Luisa Coheur, Fernando Pereira and Ben Taskar. My main research interests are Machine Learning and Natural Language Processing. My current focus in on unsupervised learning with high level supervision in the form of constraints. I am a proud member of the Spoken Language Systems Lab (L2F) in Lisbon and of the Penn Research in Machine Learning (PRiML).