POSTDOCTORAL POSITION AT LIMSI-CNRS

   Topic: New methods for learning Named Entity Recognition systems
   in a multilingual context.

The development of Natural Language Processing systems is impeded by
the absence of annotated data in some languages.  A possible solution
consists in transferring analyses available in one language to
comparable texts in another language.  This makes it possible to train
new systems based on these new annotations.

The proposed project will tackle more specifically the case of Named
Entity Recognition in a context where a 'synchronous comparable
corpus' (or 'noisy parallel corpus') is available: AFP news in French,
English, Spanish (German, Portuguese and Arabic are also available).

The goal is to take advantage of the parallelism of news written in
multiple languages to help recognize named entities:

- in a language for which a Named Entity Recognition system is
  already available, strengthen recognition through entities found in
  parallel articles in other languages;

- in a (target) language for which no NER system is available,
  transfer into that language the analyses obtained in a (source)
  language for which a NER system is available, and use them to train
  a system in that (target) language.

LIMSI has:

- NER systems for French, English and Spanish (with varying levels of
  performance);

- methods for detecting translation relations among news articles.

This work takes place in the context of the EDyLex project, funded by
the French National Research Agency, whose goal is to process unknown
words in texts (http://sites.google.com/site/projetedylex/).  The work
of the successful candidate will be focused on the detection and
categorization of named entities in a multilingual context.

The work will be performed at LIMSI-CNRS in Orsay
(http://www.limsi.fr/) on the campus of University Paris-South.  Two
teams at LIMSI are involved in the project: ILES (written and signed
language processing) and TLP (spoken language processing).


QUALIFICATIONS AND POSITION

The successful candidate will have a track record of Machine Learning
for Natural Language Processing research.  A strong preference will be
given to those candidates with experience in Named Entity Recognition
or Speech Language Understanding.  Fluency in one or more languages of
the project beyond English is mandatory.  Applicants should have
received (by the starting date) a PhD in Machine Learning,
Computational Linguistics or related areas.

This position is for 12 months and may begin as early as Oct 1st,
2011, or soon thereafter.  Salary follows CNRS scales and depends on
the candidate's experience (the minimum monthly net salary is about
2,000 €).

To apply, please send a cover letter, describing how the applicant's
knowledge and research background will contribute to the project, a
CV, and the names and contact information of two referees to:

 Pierre Zweigenbaum (pz@limsi.fr) and Sophie Rosset (rosset@limsi.fr)