Title: Representing Human Language: a Continuous Space Approach

Speaker: Professor Mari Ostendorf
                 University of Washington

Date: 3 April 2013
Time: 2:00pm – 3:30am 
Venue: Rm 513, William M. W. Mong Engineering Building, CUHK

Abstract:

Traditionally, most work in statistical language processing has treated words as categorical variables and represented probabilities of words with non-parametric distributions. For applications where a relatively large amount of transcribed text is available, such models can be very effective. However, a high-dimensional non-parametric model is not well suited to leveraging regularities in language in limited training conditions, such as for domain adaptation or low resource languages, or highly inflective languages. Parametric continuous-space models are better suited to such challenges, which motivates the problem of finding mappings of words to a continuous space. In this talk, we survey prior work in continuous-space modeling of language, including latent semantic analysis and neural network models, and introduce a new approach within the context of an exponential model that treats unseen events as a regularization problem. Neighbors in the continuous space for a sequence model correspond roughly to syntactic/semantic categories. Key to the effectiveness of this approach is combining the smooth distribution with a sparse discrete model of exceptions. Inspection of the sparse component provides insights into the idiosyncracies of speakers and speaking style.

About the speaker:

Mari Ostendorf is a Professor of Electrical Engineering at the University of Washington. After receiving her PhD in electrical engineering from Stanford University, she worked at BBN Laboratories, then Boston University, and then joined the University of Washington (UW) in 1999. At UW, she is an Endowed Professor of System Design Methodologies in Electrical Engineering and an Adjunct Professor in Computer Science and Engineering and in Linguistics. From 2010-2012, she served as the Associate Dean for Research and Graduate Studies in the College of Engineering. She has previously been a visiting researcher at the ATR Interpreting Telecommunications Laboratory and at the University of Karlsruhe, and a Scottish Informatics and Computer Science Alliance Distinguished Visiting Fellow. Currently, she is an Australia-America Fulbright Scholar at Macquarie University. Prof. Ostendorf's research interests are in dynamic and linguistically-motivated statistical models for speech and language processing. Her work has resulted in over 200 publications and 2 paper awards. Prof. Ostendorf has served as co-Editor of Computer Speech and Language, as the Editor-in-Chief of the IEEE Transactions on Audio, Speech and Language Processing, and she is currently the VP Publications for the IEEE Signal Processing Society. She is also a member of the ISCA Advisory Council. She is a Fellow of IEEE and ISCA, a recipient of the 2010 IEEE HP Harriett B. Rigas Award, and a 2013 IEEE Signal Processing Society Distinguished Lecturer.