May and June 2019 marked (yet) another record month for Neural Machine Translation (NMT)
On May 17, 2019 over 80 language industry leaders gathered at the 2019 edition of SlatorCon conference in London. Participants represented a diverse mix of industry stakeholders from end buyers, vendor CEOs, analysts, and investment bankers involved in industry M&A. Delegate profiles cut across the spectrum from pharma to game and media localization, digital marketing and more, with many traveling in from outside of the UK.
Currently, one of the fastest growing industry verticals (and one that is still relatively untouched by NMT) is media localization. As ZOO Digital CEO Stuart Green confirmed in his presentation, “demand is outstripping supply”. His company has developed a cloud-based media localization platform, which is challenging the old ways of providing services such as subtitling, dubbing, and voiceover. Dubbing, for example, would traditionally involve bricks-and-mortar operations and require the voice artist to physically be present at a studio. ZOO has struck big over the past two years as Netflix and other over-the-top content (OTT) providers expanded so aggressively across many international markets that the existing media localization infrastructure could barely cope. The result was exceptional revenue growth and a skyrocketing share price.
Michaela Bartelt-Krantz, Senior Localization Director at game giant Electronic Arts runs EA’s internal language service, which works mostly directly with freelancers and single language providers (SLVs) and has its own dedicated engineering team. Bartelt walked participants through their operating model (more to come in a separate article) and closed on what she called a “science fiction vision” of real-time localization powered by machine translation, voice generation, and automated testing.
Naturally, Jean Senellart, CTO of machine translation provider SYSTRAN, stayed on the subject of advanced technologies when he delivered his talk about how open source is rapidly changing the language industry’s most transformative technology: neural machine translation.
Since its emergence around three years ago, NMT has already undergone three major technological shifts from recurrent neural networks, to convolutional neural networks, and now attention-based neural networks. Senellart said the underlying driver for big tech’s quest to dominate in NMT is the fact that NMT is “the gateway for all natural language processing (NLP) technologies” and highlighted that the “real battle is on the AI framework and computation backend”.
Hogarth’s Global Head of Language Services, Romina Franceschina, explained why some of her clients prefer to have linguists and project managers onsite: “They are very passionate about meeting the people that will carry their brand in their language”.
Neural machine translation (yet again) and AI more broadly were a hot topic in the panel. The discussion concluded with a lively exchange on the possibility of natural voice generation creating all sorts of moral dilemmas.