Post-doc at NRC/Interactive Language Technology

[Cyril Goutte]

Statistical Document Modelling for Improving Translation Workflows


The Interactive Language Technology (ILT) group of the National
Research Council Canada is looking for a candidate to fill a 2-year
researcher position in the area of Machine Learning applied to Natural
Language Processing. The group is located in Gatineau, Quebec (one
bridge away from Ottawa).


Your Challenge:
The researcher will initially work on a new collaboration with a large
government agency. The goal of the project is to investigate the use
of unsupervised or lightly-supervised techniques to uncover some
useful implicit structure from large, unstructured, multilingual
collections of documents, in order to improve the efficiency of
computer-aided translation tools such as statistical machine
translation, translation memories and terminology support tools.

Researchers from the ILT group have been successfully exploiting
the framework of multiview learning [4-5] in order to design novel
techniques for semi-supervised document classification and ranking
[1-3]. The post-doc will build on this work and explore
the theoretical foundations of lightly-supervised, multilingual
document organization models, develop advanced statistical models, and
test them on large collections of multilingual document.


Responsibilities:
- Build on Machine Learning theory to characterize and improve
lightly-supervised techniques for organizing large collections of
multilingual documents;
- Design and implement advanced statistical models;
- Build prototypes, and test them on large multilingual document
collections;
- Publish in leading ML/IR conferences and journals.


Requirements:
- PhD in Statistics, Computer Science or Natural Language Processing,
with experience in the areas of textual document modelling and
organization, and/or statistical machine translation;
- Publication record in relevant conferences and journals;
- Proven ability to prototype and run large-scale experiments;
- Ability to work in English and/or French.


Duration:
2 years.

Start date:
Early September 2010

Salary:
Salary will be commensurate with experience, on a scale starting at
60,940 CAD, plus full benefit package available to NRC researchers.

Informal inquiries are welcome and should be directed to
Cyril....@nrc.ca or Massih-R...@nrc.ca.
Applications should be made through the NRC Research Associate
Program:
http://www.nrc-cnrc.gc.ca/eng/careers/programs/research-associate.html


Background:
The National Research Council Canada (NRC) is the Government of
Canada's premier organization for research and development.  NRC is
composed of over 20 institutes and national programs spanning a wide
variety of disciplines. 

The Interactive Language Technologies (ILT) Group develops new and
innovative technologies to enable assisted translation of texts
written in free-form human languages and to support the creation and
management of multilingual terminology and content.


References:
[1] M.-R. Amini, C. Goutte, A Co-classification Approach to Learning
from Multilingual Corpora. Machine Learning Journal, 79(1-2):105-121,
2010

[2] M.-R. Amini, C. Goutte, N. Usunier, Combining Coregularization and
Consensus-based Self-Training for Multilingual Text
Categorization. Proceedings of the 33rd Annual ACM SIGIR Conference
(SIGIR 2010).

[3] M.-R. Amini, N. Usunier, G. Goutte, Learning from Multiple
Partially Observed Views -- an Application to Multilingual Text
Categorization. Advances in Neural Information Processing Systems 22
(NIPS 2009) pp. 28-36

[4] K. Crammer,M. Kearns, and J.Wortman. Learning from Multiple
Sources. Journal of Machine Learning Research, 9:1757-1774, 2008.

[5] D. S. Rosenberg and P. L. Bartlett. The Rademacher Complexity of
Co-regularized Kernel Classes. Proceedings of Artificial Intelligence
& Statistics, (AISTAT 2007) pp. 396-403, 2007.