October 10, 2006 07:30 ET

AppTek Scores Best on "Noisy Data" Under NIST Machine Translation MT06 Evaluation

MCLEAN, VA -- (MARKET WIRE) -- October 10, 2006 --AppTek's machine translation system scored best on "noisy data" (i.e. text from speech input, newsgroups, etc. as opposed to "clean data" like newswire text) in the most recent National Institute of Standards & Technology evaluation, according to the test results published in the September 2006 workshop. The Hybrid Machine Translation system tested, TranSphere HMT, was a culmination of a hybridization effort by AppTek to bring together both of its platforms under one application. AppTek's scientists utilized its rich rule-based engine to augment & enhance the newly acquired statistical MT platform, pushing the state-of-the-art in Machine Translation design to the next level.

AppTek plans to adapt the hybrid approach to cover all of its language pairs, including bi-directional English with Arabic, as well as Persian, Pashto, Indonesian, Tagalog, Turkish, Chinese, and Korean into English. The next phase will be to adapt it to bi-directional English with German, Spanish, French, Polish, Italian, Russian, Portuguese, Ukrainian, Hebrew, and Dutch.

Hassan Sawaf, AppTek's R&D Chief, presented the results at NIST's workshop on September 7. He stated: "The results of the NIST 2006 evaluation are very satisfying for AppTek. They show the strength of the new hybrid MT approach by AppTek. Despite the fact that this is the first time AppTek participated in the NIST MT evaluation, we scored very high, especially on noisy data. This verifies that the hybrid MT beats both pure rule-based MT as well as pure statistical MT approaches." The full script of his presentation is available on

NIST will publish the results on their website ( NIST conducts these evaluations in order to support machine translation (MT) research and help advance the state-of-the-art in MT technology, rather than as a competition. As such, the results are not to be construed or represented as endorsements of any participant's system or commercial product, or taken as official findings on the part of NIST or the U.S. government.

Contact Information