According to DeepL`s press paper, the system runs on the world`s 23rd largest supercomputer in Iceland (for cheap electricity), where it can translate a million words in less than a second and reach BLUE RECORD values. And DeepL is not finished. He also claims that its “revolutionary neural architecture” makes it “the most accurate and natural machine translation tool.” But here`s the caveat: in long sentences, DeepL and Google Translate both collapse and produce something that`s almost impossible to understand as a full-fledged German phrase without reading the English original. That is, NMT is known to deal with longer sentences. Slator conducted his own, totally unsympathific experiment and pitted DeepL against Google Translate using three short paragraphs of a Bloomberg article about an antitrust investigation against Apple in China. Located in the beautiful maritime city of Zurich, Switzerland. Barely two years after printing on the translation tech scene, neural machine translation (NMT) is all the TM community is talking about. Microsoft, Google, Facebook and other big tech companies have all switched to NMT, as have the European Patent Office and the World Intellectual Property Organization. Even end buyers are starting to build their own systems from open source models. Not that other TM suppliers tend to overly modesty when introducing their products, but the Germany-based DeepL comes out of the door with flaming rifles. Visit us at SlatorCon New York on the 12. October 2017 and receive an update on the state of the art in machine neural translation from one of the leading researchers in this field, Kyunghyun Cho, Assistant Professor of Computer Science and Data Science at New York University.
. . .