Translation Tech in 2015
Beyond continued fragmentation of languages on the internet driven by internet/mobile growth —now 80% of users do not understand English — 2015 was an amazing year for machine learning. The revolution is supposedly coming to NLP next. It is already impacting speech recognition.
Deep Learning x Translation?
But natural human language is complex —I share the view that the hardest problems will take more than deep learning (brute force)   . So despite 2015 being the year of intelligent virtual assistants from x.ai to Facebook M to Viv Labs to Microsoft Cortana for iOS and Android, the tech in prod still relies on a mix of humans and hand-coded templates.
The Big Players
Microsoft rolled out real-time voice-to-voice translation in Skype.
In Q1, Twitter officially re-launched integrated translation, similar to that of Facebook and Google+. (Twitter had unlaunched the feature years ago because of quality problems.) Twitter translation, like that of Facebook and many others, is powered by Microsoft.
Android made it easier for apps to integrate Google Translate. And Google Translate rolled out a smoother translation of speech and images. The growing importance of mobile input methods drove their 2014 acquisition of Word Lens and these latest launches.
It’s subtle but very important — informal and dirty data are the reality. When this feature comes to mobile, and for all languages, the impact will be massive.
Translation Startups and Acquisitions
These echoed eBay’s aquisition of AppTek’s Germany-based translation team a year before. Then Unbabel launched more intelligent crowdsourcing.
To end the year, Lilt, another startup founded by former Google Translate engineers, launched a new and amazing experience for professional translators.