SYSTRAN remporte une première place pour la qualité de ses traductions à la compétition internationale « WMT »

SYSTRAN’s ground breaking hybrid machine translation engine outperforms Google and others in English to French translation

PARIS, FRANCE and SAN DIEGO, CA – May 28, 2009
SYSTRAN, the leading provider of language translation technologies, today announced that it ranked first in an English to French translation task at the fourth annual Workshop on Statistical Machine Translation in Athens, Greece. This workshop, which was part of the EACL 2009 Conference, was organized by Philipp Koehn, PhD, a recognized scientist in statistical machine translation. The goal of the competition was to compare the quality of multiple machine translation systems for 10 European language pairs, and to set up objective evaluation criteria for machine translation quality.

SYSTRAN’s latest innovation, a hybrid approach to machine translation, outperformed Google Translation, academic statistical machine translation (MT) systems and several rule-based systems in the category of English to French News Translation. Making smart use of a reasonable size training corpus SYSTRAN delivered more accurate translation than very large systems like Google, which rely on a gigantic language model that is based on a Web index. SYSTRAN’s hybrid MT engine will be officially released in Enterprise Server 7 on June 3rd.

SYSTRAN Enterprise Server 7, the first hybrid machine translation (MT) product designed for enterprise use, is based on a combination of self-learning and linguistic technologies. SYSTRAN 7 delivers highly consistent and quality translations, while significantly reducing the cost and time to customize the software.

“SYSTRAN’s unique hybrid technology, which uses both deep linguistic analysis and self-learning, outperforms any other approach,” said Dimitris Sabatakakis, CEO of SYSTRAN. “Our hybrid technology is designed to learn from fewer resources and is perfectly suited for high-quality specialized translation. As we have done for many years, SYSTRAN continues to be the leading innovator in the machine translation market.”

Philipp Koehn, a statistical machine translation expert and the workshop organizer said “this workshop demonstrated that the rule-based approach combined with statistical and other techniques lead to highly competitive MT systems.”

Société

SYSTRAN est le leader mondial des technologies de traduction automatique et commercialise des solutions et produits innovants pour les postes individuels, les applications d'entreprise et les services en ligne qui facilitent la communication multilingue grâce à 52 paires de langues et 20 dictionnaires-métiers.

Avec plus de 40 années de recherche et développement, SYSTRAN compte parmi ses clients des sociétés multinationales, les grands acteurs Internet comme Yahoo!, Altavista ou Apple et des organisations publiques comme l'US Intelligence Community et la Commission européenne.

Les produits et solutions SYSTRAN améliorent la qualité de la communication multilingue et la productivité dans des contextes professionnels aussi variés que les relations B2B et B2C, en fournissant des traductions instantanées pour les moteurs de recherche, la gestion de contenu, le support client, le commerce électronique et les échanges collaboratifs intra et inter entreprises

SYSTRAN a son siège social à Paris (France) et une filiale à San Diego en Californie (Etats-Unis).

SYSTRAN (Code ISIN FR0004109197, Bloomberg : SYST NM, Reuters: SYTN.LN) SYTN.LN) est cotée sur le compartiment C d'Euronext Paris.

Contact

Contact Presse
Thomas Boullonnois
781-684-0770
Email : thomas@rumeurpublique.fr

Veuillez compléter l'information ci-dessous pour télécharger le document.

*Champs obligatoires

Des informations n'ont pas été correctement renseignées. Veuillez compléter les champs encadrés en rouge

Informations société
Information contact
Code de sécurité

Saisissez les caractères comme visibles dans l'image ci-dessus (sans différentiation de casse). Veuillez cliquer sur l'image si vous avez des difficultés pour lire ces caractères.