Corpora: ACL-2001 Workshop on Data-driven Call for Papers

From: Priscilla Rasmussen (rasmusse@cs.rutgers.edu)
Date: Wed Feb 21 2001 - 00:29:09 MET

  • Next message: Gabriela Cavaglia: "Corpora: Italian corpus"

     
             Call for papers
             Workshop on Data-driven MT
             ACL'2001 Conference
             Toulouse, France
             Invited speaker: Hermann Ney, RWTH Aachen
             Deadline for paper submissions: April 6, 2001
             Deadline for notification of paper acceptance: April 27, 2001
             Deadline for camera-ready papers: May 16, 2001
                  
             Workshop Date: July 7, 2001
     
             Details on submissions listed below.
     
             With the increased availability of online corpora, data-driven
     approaches have become central to the NL community. A variety of
     data-driven approaches have been used to help build Machine Translation
     systems -- example-based, statistical MT, and other machine learning
     approaches - and there are all sorts of possibilities for hybrid systems.
     We wish to bring together proponents of as many techniques as possible to
     engage in a discussion of which combinations will yield maximal success in
     translation.
     
             We propose to center the workshop on Data Driven MT, by which we
     mean all approaches which develop algorithms and programs to exploit data
     in the development of MT, primarily the use of large bilingual corpora
     created by human translators, and serving as a source of training data for
     MT systems. We are specifically interested in papers about
     
                     * statistical machine translation (modeling, training,
     search)
                     * machine-learning in translation
                     * example-based machine translation
                     * acquisition of multilingual training data
                     * evaluation of data driven methods (also with
     rule-based methods)
                     * combination of various translation systems;
     integration of classical rule-based and data driven approaches
                     * word/sentence alignment methods
     
             An especially important question that we wish to address is which
     techniques are best for each of the subparts of a complete MT system -
     e.g. learning grammars, building lexicons, parsing input data,
     determining transfer principles, generating target text, etc.
     
             We will strongly encourage papers on systems which show
     demonstrable progress over previously chosen methods, and which have been
     integrated in an actual end-to-end system. Test results or demos will be
     given strongest preference for participation.
     
             Organizers:
             Jessie Pinkham, Microsoft Research jessiep@microsoft.com
     <mailto:jessiep@microsoft.com
             http://research.microsoft.com/~jessiep/
     
             Kevin Knight USC/ISI; knight@isi.edu <mailto:knight@isi.edu
             Web page http://www.isi.edu/~knight/
     
             Franz Josef Och, RWTH Aachen; och@informatik.rwth-aachen.de
             http://www-i6.informatik.rwth-aachen.de/Colleagues/och/
     
             SUBMISSION FORMAT AND INSTRUCTIONS:
             Electronic submissions only; send the postscript or pdf form of your
     submission to: Deborah Coughlin deborahc@microsoft.com .
     
             Submissions should follow the two-column format of ACL proceedings
     and should not exceed eight (8) pages, including references. We
             strongly recommend the use of ACL LaTeX style files or Microsoft
     Word Style files tailored for this year's conference. They are
             available from the ACL-2001 program committee Web-site at
     <http://acl2001.dfki.de/style/ .
     
             As reviewing will be blind, a separate identification page must be
     sent by email. The identification page should include the paper title,
             authors' names, affiliations, and email addresses, up to 5 keywords
     specifying the subject area, and a short summary (up to 5 lines).
             The paper should not include the authors' names and affiliations.
     



    This archive was generated by hypermail 2b29 : Wed Feb 21 2001 - 09:05:18 MET