Informatique et Linguistique : rencontre pluridisciplinaire

English version

Version Française


Université Paris 1 Panthéon-Sorbonne

 

Colloquium

"NooJ tools for Machine Translation :

What is next ?"

NooJ_Trad 2006

September 29th, 2006

Université Paris 1 Panthéon-Sorbonne

Paris, France

___________________________________________________

Organised by

Odile Piton (Laboratoire Marin Mersenne, Université Paris 1 Panthéon-Sorbonne)

Max Silberztein (LASELDI, Université de Franche-Comté)

Today, industrial and large-audience Machine Translation software are still producing poor quality results. Researchers in Linguistics and Computational Linguistics, as well as researchers who process multilingual Corpora, need a tool allowing them to formalize necessary linguistic resources (e.g. bilingual dictionaries and grammars), study how linguistic data should be transfered from one language to another, and simulate the behavior of MT software.

Beyond word-by-word translation, and even if we take into account compound words (pomme de terre/potato) and frozen expressions (casser sa pipe/kick the bucket), a large number of semi-frozen expressions need to be formalized which translation might not be a direct one. For instance, the French sentence "Luc a faim", which does not translate directly in English, could first be rewritten as "Luc est affamé" in order to ultimately be translated by "Luc is hungry".

NooJ, which is an open linguistic development platform, includes several mechanisms that allow it to perform transformational operations on texts. For instance, its morphological and syntactic engines are fully integrated. These mechanisms are used to produce paraphrases (e.g. to nominalize an expression), but could also be used for MT, to compare language variants, or to parse multilingual corpora.

This workshop will offer researchers and students interested by the architecture, technical aspects and linguistic description for MT software, an opportunity to meet and explore in what conditions, and with what types of modifications, NooJ could be used for MT.

For this occasion, the speakers will present convergent problematics : the development of an MT (English/Portuguese) application with NooJ, the use of a corpus processing system, and a corpus-based comparative study between two languages (classical and medieval latin).

Speakers :

Max Silberztein (LASELDI, Université de Franche-Comté)

Monique Goullet (Lamop CNRS, Université Paris 1 Panthéon-Sorbonne)

Jean-Philippe Genet (Lamop CNRS, Université Paris 1 Panthéon-Sorbonne)

Sylvie Auffret-Pignot (CRLV, Université Paris 1 Panthéon-Sorbonne)

Xavier Blanco (Université Autonome de Barcelone)

Svetla Koeva (DCL, Institute for Bulgarian language)

Tamás Váradi (Académie des Sciences de Hongrie)

Anabela Barreiro (New-York University, Université de Porto)

tutorial session "NooJ for MT", animated by

Slim Mesfar (Université de Franche-Comté )

Poster session

PROGRAMME

Monique GOULLET (Lamop CNRS/ Paris I)

Traiter automatiquement le latin : du latin classique au latin médiéval.

Jean-Philippe GENET (Lamop CNRS/ Paris I)

Le plurilinguisme médiéval et la naissance de la langue politique en Europe : le cas anglais.

Sylvie AUFFRET-PIGNOT (SELVA Paris1)

Comparaison entre l’anglais du 17ème et l’anglais moderne, à travers les récits de trois voyageurs Britanniques. ppt

Max SILBERZSTEIN (Laseldi U. de Franche-Comté)

NooJ v1.30 et la traduction automatique.ppt

Xavier BLANCO (Université de Barcelone)

Classes, Categories and Structures for Machine Translation. ppt

Svetla KOEVA (Université de Sofia)

Dictionaries in Machine Translation.

Tamás VÁRADI (Ac. des Sciences Budapest)

Machine Translation with local grammars.ppt

Anabela BARREIRO (U. de Porto, U. de New-York)

Portuguese to English translation system using Nooj. ppt

Slim Mesfar (U. de Franche Comté)

Tutoriel NooJ pour la traduction. pdf

Pour toute information sur NooJ