R&D leader

Transclick's R&D Initiatives

Transclick's mission is to be the world's leader in real-time translation of instant messaging, SMS translation, email translation and text to voice translation as well as speech to speech translation. This will require a significant investment in R&D. Fortunately, Transclick's R&D initiatives are led by a talented group of scientists specializing in artificial intelligence and machine translation.

Transclick's Chief Scientist, Dr. Leslie Barrett, leads Transclick's research efforts with VP Technology Keith Regli and Dr. Michael Dilinger, supported by our Director of Systems Engineering, Arthur Bunger. Dr. Edward Hovy, http://www.isi.edu/natural-language/people/hovy.html International Association of Machine Translation (IAMT) is Chairman of Transclick's Board of Technology Advisors.

Dr. Barrett, a Ph.D. in computational linguistics from New York University, was the principal investigator for Transclick's Small Business Innovation Research (SBIR) Grant with the United States Army Research Laboratory. After 9/11, Transclick was asked to create a prototype for real-time language translation of Arabic speech and converting it to English speech in a few seconds.

Dr. Barrett of Transclick is chairing a workshop of the AMTA-2004, the 6th Conference of the Association for Machine Translation in the Americas on October 2, 2004:http://www.amtaweb.org/AMTA2004/workshop.html

On September 27, 2003, Dr. Barrett organized a workshop and delivered an important paper on systematizing MT Evaluation at the Machine Translation Summit XI in New Orleans: http://www.issco.unige.ch/projects/isle/MTE-at-MTS9.html

September 9, 2003: Dr. Leslie Barrett presented a paper on machine translation evaluation techniques and compares quality of MT to human judgments: http://nlp.cs.nyu.edu/nycnlp/abstract09-12-2003-talk1.shtml

Machine translation can be evaluated using the well-known measures precision, recall, and the F-measure. The F-measure has significantly higher correlation with human judgments than recently proposed alternatives. More importantly, the standard measures have an intuitive graphical interpretation, which can facilitate insight into how MT systems might be improved. The relevant software is publicly available fromhttp://nlp.cs.nyu.edu/GTM/

However, evaluation of MT evaluation measures is limited by inconsistent human judgment data. Moreover, even though human evaluation of MT is itself inconsistent and not very reliable, automatic MT evaluation measures are even less reliable. Transclick has patents pending to increase the level of input from human judgment through context recognition of topics and switching to the dictionary in real-time that optimizes the real-time translation.

Dr. Barrett published a paper on natural language processing ontologies with Anthony R. Davis, Ph.D.: http://www.bultreebank.org/OntoLex02/OntoLex02Paper02.pdf