Date: Monday 11/5/2012 
Time: 2:15 pm 
Speaker: Qun Liu (Dublin City University)
Venue: CUNY Graduate Center Science Center (Room 4102) 
Title:  Context-Aware Rule-Selection for Statistical Machine Translation

Abstract:

All translation models use some kinds of translation rules to map a source
word, phrase or structure to a target one.  In current translation models the
probability of this mapping are static and unaware of context, which means the
rule selection in decoding is actually made by the language model or
re-ordering model.  In this talk we proposed an idea to incorporate context
features into translation models to conduct better rule selection.  This idea
is compatible with the mainstream linear framework for statistical machine
translation and is very suitable to incorporate various context features in a
local manner without influence on the global model.  We successfully implements
this idea in various translation models using different context features and
obtain significant improvements over baseline systems.

Bio:

Prof. Qun Liu is Professor of Machine Translation at Centre for Next Generation
Localisation (CNGL) and in the School of Computing at Dublin City University
(DCU).  Before joining CNGL, Prof. Liu was a Professor and the Director of the
Natural Language Processing Research Group at the Institute of Computing
Technology (ICT) at Chinese Academy of Sciences (CAS) in Beijing.  He got his
Master degree in ICT at CAS in 1992 and his Ph.D. degree in Peking University
in 2004.  His research interests include Machine Translation, Human Language
Technologies and Natural Language Processing for Chinese.  He is the author or
co-author of more than 150 research publications in the areas of Machine
Translation and Natural Language Processing, including many publications at the
top international ACL, COLING, EMNLP, IJCNLP and AAAI conferences.  His recent
researches mainly focus on linguistically enhanced statistical machine
translation models and evaluation approaches.