[Colloquium] TTIC Talks: Kevin Gimpel, CMU

[Liv Leader]

REMINDER:

When:     Thursday, February 16 @ 1 p.m.

Where:   TTIC Conference Room #526, 6045 S. Kenwood Avenue, 5th Floor

Who:      Kevin Gimpel, CMU

Title:       Statistical Modeling and Learning for Machine Translation

Recent years have seen a flurry of research in translation modeling for
statistical machine translation. Widely-used approaches include (1) models
based on flat phrase-to-phrase mappings, and (2) models that use syntactic
structure. I will present an approach that combines the strengths of these
two in a single model. Experiments show that it leads to improved
translation quality over state-of-the-art systems. When supervised
syntactic parsers are not available, I will show that unsupervised parsers
can be substituted with minimal loss in translation quality. I will also
discuss the problem of training model parameters. An abundance of
techniques has been developed by the statistical machine learning
community, but the machine translation problem is different from other
structure prediction tasks in certain ways. These differences have caused
well-known learning algorithms to lose theoretical guarantees when applied
to machine translation. I will propose a novel algorithm that does offer
theoretical guarantees and shows empirical advantage over state-of-the-art
approaches.

Host: Karen Livescu, klivescu at ttic.edu

-- 
Liv Leader
Human Resources Coordinator

Toyota Technological Institute Chicago
6045 S Kenwood Ave
Chicago, IL 60637
Phone- (773) 702-5033
Fax-     (773) 834-9881
Email-  lleader at ttic.edu <jam at ttic.edu>
Web-   www.ttic.edu
<http://www.ttic.edu/>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.cs.uchicago.edu/pipermail/colloquium/attachments/20120215/efed75bf/attachment.htm