Dr. Ariadna Quattoni, postdoc at the LARCA group, together with Drs. Xavier Carreras and Lluis Marquez of Dept LSI-UPC, are the recipients of one of the Google Research Awards 2010, in the Natural Language Procesing category.

Their project, Semi-supervised Syntactic Parsing by Inferring Hidden Sequences, presents a novel method for semi-supervised dependency parsing that is able to leverage large amounts of unlabeled data in addition to labeled data. As in previous work, the main idea is to use the unlabeled data to learn intrinsic features that can improve the results of a standard discriminative parser. In contrast with previous work on semi-supervised dependency parsing, they propose to learn hidden features derived from correlations of words with respect to their syntactic behaviours. This will be achieved by applying novel methods of inducing automata using spectral techniques.

The awards, besides funding to push the project forward, implies the appointment of a Google sponsor, with whom recipients can discuss research directions, provide progress updates, engage in knowledge transfer, etc. Google retains no intellectual rights from the research, and in fact encourages restults to be open-source and widely published.