What is Real-Time Machine Translation?

Real-time machine translation refers to the process of a computer producing real-time translation between human languages at a rate and quality level that would allow it to be used for speech-to-speech, speech-to-text, text-to-speech, or text-to-text applications. Two early examples of where this technology is heading can be found in Siri, Apple's personal assistant, which listens to, transcribes, and acts upon speech, and Google Translate, which rapidly translates to a reasonable standard between over 50 languages. While some may argue that these examples are evidence that computers have not yet mastered this activity, recent developments in real-time machine translation are bringing us closer to smarter, faster, more accurate, and more culturally aware systems for communicating across language barriers. Systems that can listen to student speech, then coach (or rate) it for pacing, tone, dialect, and accuracy of pronunciation have been in production for some time. In the next generation of machine translation, these sorts of technologies will merge to provide tools that can deliver more accurate real-time translations, and render them into speech that includes fine-grained nuances of pronunciation, tone, and more. Ray Kurzweil, a major thought leader in the area of machine translation, believes that before 2030, machines will reach a sufficient level of understanding of human written and spoken communications to allow for routine, everyday, seamless and highly accurate translation. Many believe the technology will enter specialized uses much sooner, with a great many applications in learning, teaching, and global communications.