Undetectable Humanizer: Lifetime Subscription
Transform AI-Generated Text into Human-Like, High-Ranking Content & Bypass Even the Most Sophisticated AI Detectors
Get 95% Deal

Microsoft demos Skype real-time translator

Martin Brinkmann
May 28, 2014
Updated • May 28, 2014
Microsoft
|
1

Machine translation has come a long way in the past decade and while research and projects happened mostly behind closed doors or in academia, end user products have gotten better as well.

Web browsers for instance support a variety of translation options, from the built-in Google Translate in Chrome to dozens of add-ons for Firefox that provide you with translation options.

There have been attempts to translate chat in real-time, Clownfish and Skype Chat Translator are two plugins for Skype that offered those features.

But real-time voice translation, that's is new, at least when it comes to end-user products.

Microsoft has demoed a machine translation feature for Skype yesterday that showed how far research has already progressed in recent time.

You see a communication between the English speaking Pall and the German speaking Diana Heinrichs that is translated in real-time by the Skype application.

You should not expect a perfect quality just yet, and there are pauses as well before the computer starts to speak what is has translated, but all in all, it is astonishing how well this works.

As far as quality of the translation is concerned, it is comparable to Bing Translate. While that is sufficient to understand the meaning of what is being said most of the time, words are sometimes used out of context. A simple example: Pall's question "Hello Diana, how are you doing" is translated to "Hallo Diana, wie dir's geht" which is not a proper sentence in German. While the meaning is clear, it should have translated it to "Hallo Diana, wie geht es dir".

It is the equivalent of saying "Hello Diana, how doing are you" in English which is not proper as well even though the words are the same.

There is a second longer sentence where Pall answers a question about moving to London where the translation engine is using a wrong word.

The answer "Yes, I'm gonna be planning to move there sometime between the middle of June to the beginning of July if everything goes, but track" is translated to "Ja ich plane derzeit irgendwann zwischen Mitte Juni bis Anfang Juli zu verschieben, wenn alles geht, aber verfolgen".

The main issue is the word verschieben. While move can mean verschieben in German, in this particular case it should have been translated with umziehen instead as verschieben ist not used in this context.

The translations work well for the most part. In the particular case of translating English to German, it seems to use improper sentence structure at times.

"I think my family would love it there this Summer, but I'm sure my kids will miss their friends back in Seattle" is translated to "Ich denke, dass die Familie es dort diesen Sommer lieben würde, aber ich bin sicher, dass meine Kinder vermissen ihre Freunde zurück in Seattle" is another example.

There is nothing wrong with the first part of the translation, but in the second part, it is again the verb that is placed in the wrong location. The proper translation should have been "dass meine Kinder ihre Freunde in Seattle vermissen würden".

If you speak German, you can still understand what is being said though and that is probably more important at this point than proper sentence structure or use of words.

Here is Microsoft's promo video for the new feature.

Microsoft will launch Skype Translator as a beta feature for Skype on Windows later this year.

Advertisement

Tutorials & Tips


Previous Post: «
Next Post: «

Comments

  1. Shai said on June 1, 2014 at 4:54 pm
    Reply

    I can go on and on about Machine Translation and the hoax that it is, but for the sake of brevity I will just say this: Translation is all about language, knowledge and skills, and those language and knowledge are all about communication. Language is not data, or a big data, problem that can be solved by crunching large data-sets and applying “clever” algorithms. Therefore, MT is not even translation in the true meaning of the word. It is a statistical language mapping and transformation algorithm, mistakenly dubbed as “translation” for commercial purposes.

    The notion that MT has come a long way is a little misguided. The major advancement in generic MT is that it became more available to the general public, and this is more the result of the developments in hardware and change in use patterns (moving to the cloud).

    I wouldn’t call this presentation or technology magical, not by any stretch of the imagination. The real-world applications of MT are far more limited than what the proprietors and promoters of the technology claim. The damages of poor translation quality, however, are very real (including sending sensitive communications to the servers to be “translated”).

Leave a Reply

Check the box to consent to your data being stored in line with the guidelines set out in our privacy policy

We love comments and welcome thoughtful and civilized discussion. Rudeness and personal attacks will not be tolerated. Please stay on-topic.
Please note that your comment may not appear immediately after you post it.