Two young girls using Skype Translator

Skype Translator offers the stuff of sci-fi: real-time (or as close as possible to real-time) translations of a voice conversation in two different languages. When one user speaks, a computerized voice speaks the translation to the other user, and vice versa. A running transcription of the original and translated words appears in a sidebar, ensuring that any misunderstanding from the machine is obvious to the speaker.

Microsoft has released Skype Translator Preview to select invitees on Windows 8.1. For now, the Preview can handle English and Spanish for voice translations and more than 40 languages for IM translations.

Microsoft has been developing the technology behind the Preview for over 10 years, which relies on machine learning to know the difference between words like “not” and “knot”. This is no small task, as context clues are sometimes the only way to tell, and understanding the context of a conversation is a sophistication in artificial intelligence that has not existed before. Translating written words is fairly straightforward (minus spelling errors), but speech is chock-full of stutters, pauses, slang, and fillers like “um”. In addition, some languages order subject-verb-object differently, and the program has to understand not only what a word means but how to construct translated sentences.

Even with all the technical challenges, no one doubted on-the-fly speech-to-speech translation was possible; it was merely a question of when. The answer is, apparently, now.

Watch the Translator video to see it in action between classrooms far removed.