Pattern REcognition-based Statistically Enhanced Machine Translation

  • Dr. George  Tambouratzis graduated from the National Technical University of Athens (NTUA) in 1989. He received his M.Sc. in 1990 and his Ph.D. in 1993 both from the Department of Electrical Engineering of Brunel University, England. Since February 1996 he has been with ILSP and since March 1999 he has held a research post, working on the application of machine learning techniques, neural networks and evolutionary computation techniques. He has participated in several R&D projects, and has served as the co-ordinator of the IST-FP6-VEMUS and ICT-FP7-PRESEMT projects.
George Fox Lt2
Fri, 29/07/2011

The topic of Machine Translation (MT), that is, the automatic translation of a given text from a source to a target language, has been gaining importance in the current multilingual community, where vast amounts of information are available in textual format, but in a variety of languages. The PRESEMT (Pattern REcognition-based Statistically Enhanced MT) project aims to develop a flexible and adaptable MT system, based on a language-independent approach. PRESEMT spans several disciplines such as linguistic technology, corpus linguistics, parallel processing systems, machine learning and evolutionary computation.

The presentation will initially address the importance of the MT task, the shortcomings of current approaches and the benefits of the proposed approach. Then, the user groups expected to benefit the most from the project will be indicated. The key elements of the system will be presented, with emphasis on the progress expected in the main disciplines addressed by the project, as well as the method in which these results are combined. Finally, details on the project progress will be provided, together with a timetable of expected milestones and forthcoming dissemination and exploitation avenues.

PRESEMT Research Project Symposium Session (Part 1) at ECOOP 2011 from Phil Greenwood on Vimeo.

PRESEMT Research Project Symposium Session (Part 2) at ECOOP 2011 from Phil Greenwood on Vimeo.

Target Audience: 

The presentation will be of relevance to a number of different audiences:

  1. Users and potential users of machine translation software who want to learn about the latest state of the art in the domain and how this can be applied in software development and other usage scenarios;
  2. Software developers and researchers aiming to produce more natural interfaces for software systems; the translation and analysis modules developed in the PRESEMT project may provide a good basis for realising such support; and
  3. Researchers interested in object-orientation and other modularisation techniques who want to see and discuss how such techniques can be applied in a particular and complex domain.
Outline of Sessions: 
  • Morning (11am -- 12.30pm): Presentations:
    • Introduction to machine translation and PRESEMT principles (~30 min)
    • Challenges and Solutions (~60 min)
  • Afternoon (2pm -- 3pm): Hands-on exploration of core PRESEMT modules
  • 3pm -- 3.30pm: Discussions and Networking
PRESEMT.pdf275.74 KB