English and Spanish speakers can now have a Skype conversation without learning the other language or typing. Skype Translator is now available in a “preview” edition to anyone on Windows 8.1.
The service had previously been in a limited-numbers preview but has now rolled out publicly, though you will need to sign up to use it. When you use the service you can speak in English or Spanish and your words will be repeated in the other language for both parties to hear.
The original and translated text will also appear on screen as subtitles. That’s partly in case you need to review a translation because it didn’t quite make sense when you heard it, and partly so that Microsoft has a searchable archive which it uses when checking user feedback about which translations were and were not accurate.
The technology behind it is simple enough: it’s a combination of speech recognition and automated translation. The trick is carrying this out in a way that’s both fast enough for normal conversation and accurate enough to be understandable. According to Microsoft’s tech team, one of the keys to the speech recognition is identifying filler sounds like “erm” and “ah”, along with repeated words, and eliminating them before making the translation.
The set-up with both parties hearing the original and translated speaking helps slow down speech and avoid the natural inclination to speak over one another, something that makes translation far harder.
Microsoft has published a video showing two classes of schoolchildren from Tacoma and Mexico City having a conversation in their respective languages. As the video is edited, it’s hard to tell how smoothly the conversation actually went, though it’s notable that Microsoft did include a couple of clips of translations where the wording doesn’t seem quite right or natural, but the meaning can be deduced from the context.
Mashable tried its own experiment and found conversation was certainly possible, but there was some definite confusion where a word could have two meanings (one of them a name.) Perhaps unsurprisingly given the translation “vault” has been developed from real Skype conversations, it managed a perfect translation of a discussion about whether both parties could see one another on screen.