SYSTRAN Achieves High Quality Translation with Hybrid MT Engine Trained on the European Commission’s Acquis Communautaire Multilingual Corpus

PARIS, FRANCE and SAN DIEGO, CA – December 17, 2009
SYSTRAN, the leading provider of language translation technologies, today announced excellent results were achieved by training its new award-winning hybrid MT engine on the DGT Multilingual Translation Memory of the Acquis Communautaire.

The Acquis Communautaire Corpus, released by the Commission’s Joint Research Center, is the largest existing parallel corpus for 22 languages and offers the ability to create 462 language pair combinations.

The new hybrid MT engine embedded in SYSTRAN Enterprise Server 7 was released in June 2009 and is the outcome of a three-year cooperative research program between SYSTRAN and Dr. Philipp Koehn of Edinburgh University.

Dr. Koehn is the scientific lead of the Moses software, an open source statistical machine translation system that has become the reference for research in this domain.

“SYSTRAN bridges the gap between research and development in innovative MT software”, said Dr. Philipp Koehn. “SYSTRAN’s expertise on scalable software applications offers a unique opportunity for leveraging statistical machine translation technology”.

SYSTRAN’s innovative hybrid technology uniquely leverages this highly consistent corpus to achieve high quality translation. “On the basis of these excellent results, we are now working on the same approach for all European language pairs”, adds Dr. Jean Senellart, SYSTRAN’s Chief Scientist.

About SYSTRAN Enterprise 7 hybrid technology

Released in June 2009, SYSTRAN Enterprise Server 7 is a fully integrated translation server powered by SYSTRAN’s hybrid technology. Based on self-learning techniques, SYSTRAN Enterprise Server 7 can be trained on existing and validated bilingual texts to achieve cost-effective publishable quality translations. Hybrid technology combines the flexibility of statistical machine translation and the reliability of rule based linguistic approach.

About Acquis Communautaire Multilingual Corpus

The Acquis Communautaire is the entire body of European legislation, including all treaties, regulations and directives adopted by the European Union (EU), and the rulings of the European Court of Justice. […] This corpus is available online and the result of the European Commission’s effort to support multilingualism.


SYSTRAN ( is the market leading provider of language translation software products and solutions for the desktop, enterprise and Internet that facilitate communication in 52 language combinations and in 20 vertical domains. Use of SYSTRAN products and solutions enhances multilingual communication and increases user productivity and time-savings. SYSTRAN delivers real-time language solutions for search, content management, online customer support, intra or inter-company collaboration, and eCommerce. With the ability to facilitate communication in 52 language combinations and in 20 vertical domains, SYSTRAN’s software is the choice of leading global corporations, portals including Apple, Yahoo! and AltaVista, and public agencies such as the US Intelligence Community and the European Commission.

SYSTRAN has been pioneering advances in machine translation for over four decades. Its latest achievement, SYSTRAN Hybrid MT, combines the predictability and language consistency of rule-based machine translation with the fluency of statistical MT.

SYSTRAN is headquartered in Paris, France with a North American office located in San Diego, California, USA.

SYSTRAN (Code ISIN FR0004109197, Bloomberg: SYST NM, Reuters: SYTN.LN) is listed on EuroNext Paris, Compartiment C. For more information, visit

Media Contacts

Schwartz Communications
John Moran / Bill Donlan
Phone: 781-684-0770

Füllen Sie bitte das untere Informationsfeld aus, um das Dokument herunterzuladen.

* obligatorische Felder

Einige Informationen sind unvollständig. Bitte hervorgehobene Felder prüfen


Bitte geben Sie die Zeichen aus der Grafik in das Textfeld ein (Groß-/Kleinschreibung muss nicht beachtet werden). Wenn Sie die Zeichen nicht gut lesen können, klicken Sie bitte auf das Bild.