USC Viterbi alumnus Philipp Koehn is one of three scientists based in the UK to have been nominated for the European Inventor Award, which is presented annually by the European Patent Office (EPO) to outstanding innovators for their contributions to technological, social and economic progress. The award honors inventive individuals and teams whose pioneering work provides answers to modern challenges and thereby contributes to social progress, economic growth and prosperity. The 2013 winners will be announced at a ceremony in Amsterdam on May 28 in the presence of Queen Beatrix of the Netherlands.
Koehn obtained his PhD in Computer Science from the USC Viterbi School of Engineering in 2003, where he was a research assistant in the Natural Language Group at the USC Information Sciences Institute (ISI) under his thesis advisor, Professor Kevin Knight. Now a lecturer at the University of Edinburgh School of Informatics, Koehn has been nominated for the European Inventor Award for laying the foundations for modern machine translation. His revolutionary mathematical algorithms have leapfrogged the development of machine translation. Without his discovery of the phrase-base statistical machine translation (SMT) model some of the biggest names in Internet translation, including Google and Microsoft’s BING, would not be possible.
Koehn’s research on machine translation began in 1997, when he studied at the University in Erlangen, Germany, along with another household name in computer-based translation: Franz Josef Och, who now heads Google’s machine-translation team. In 2003, while working at USC ISI, the two scientists conceptualized the phrase-based SMT to reduce the restrictions of the old sluggish word-based system by translating whole sequences of words rather than one word at a time.
The basic premise of phrased-based SMT is that while a word can have several potential meanings, phrases often have only one. By developing mathematical algorithms, Koehn and Och discovered that they could statistically determine the most likely interpretation of texts. The results were astounding and together with their university professors and co-inventors, Daniel Marcu and Kevin Knight, they set about protecting their invention.
In 2002 Marcu and Knight founded the company Language Weaver. Their business model – delivering standard machine translation software – protected their patent and convinced external investors to support the company, which was sold to SDL 7 years later for $42 million.
Continuing his career at the University of Edinburgh, Koehn realized that an open source approach could allow the success of the phrase-based SMT to grow exponentially. Consequently, the open-source platform “Moses” was created, which Koehn now manages at the university.
The free, open-licensed approach means that researchers across the world can directly access and contribute to improving the system. The economic benefits are two-fold: Firstly, many companies have already integrated the free open source SMT technology into their organizations, creating a big user and client base. Secondly, a multitude of new software companies have been founded, expected to turn phrase-based SMT and related services into a several billion-dollar market in the next years.