Back during MIX09 in March, the Microsoft Translator Widget for websites was announced. People with websites or blogs can embed the Microsoft Translator Widget (powered by Microsoft Translator) into their site. Visitors of websites and blogs that have the Microsoft Translator Widget embedded can translate the content on the website without having to leave the site at all – which is awesome.
I’ve added the Microsoft Translator Widget to the Windows Experience Blog. It’s not yet deployed to all our blogs here on The Windows Blog just yet as I’d like to test it out first. The Microsoft Translator Widget will show up in the sidebar of the Windows Experience Blog where you can choose a specific language to translate the content to without leaving our site. The widget looks like this:
I would love to get some feedback on how well the Microsoft Translator Widget translates my blog posts. If you are fluent in any of the languages the Microsoft Translator Widget can translate to, translate my blog posts and let me know in the comments below how well it’s translating the blog posts.
You feedback will be much appreciated!
Please note: I am looking for feedback on how well the Microsoft Translator Widget translates content here – not feedback on the widget specifically. If you have specific feedback regarding the Microsoft Translator Widget, how it works, issues with it on your site, etc – I suggest leaving that feedback on this blog post for the Microsoft Translator Team.
If you are a website owner and wish to add the Microsoft Translator Widget to your website or blog – you can register here for a invite code.
Best website translations! www.transperfect.com/free-translations
My language is missing (hungarian), sorry MS, Google is the the king of the page translation. :D
I am brazilian-portuguese reader and it seems that the widget suffers from the same problems as the German/French guys reported.
Some words are not translated like the "I'd" that was reported and some just do not make sense.
Sometimes it looks like a misture between the portuguese from Portugal and the Brazilian one, which is not that bad but could lead to some mistakes.
The general syntax looks strange too and would need some improvements.
I would say that the page is 50% translated.
It is a good start at least!
In French, the global meaning seems mostly correct, but the sentence are gramatically uncorrect in praticaly all cases...
Suggestions for improving the translator widget for translation to German.
The German version of the translation widget is not ready for prime time. German uses a different verb word order, which the translation engine does not incorporate. For short sentences the translator seems to work quite well. The use of idioms – so common in blogs – as well as in my response here – will always be difficult. Contractions like I’ve, I’d might be added to the translator (I’ve – I have – ich habe; I’d – I would – ich möchte; but I’d – I had – ich hatte).
So first some suggestions to bloggers: Use short sentences; use commas correctly – if in doubt split the sentence into two; do not use contractions; do not use idiomatic expressions. Since neither of us can sell this, here are some suggestions for improving the translator:
For people use Leute – (Menschen means humans).
Visitors of websites and blogs that have the Microsoft Translator Widget embedded can translate the content on the website without having to leave the site at all – which is awesome.
This sentense is very poorly translated as: Besucher der Webseiten und Blogs, die der Microsoft-Konvertierung Widget embedded kann übersetzen die Inhalte der Website ohne die Website überhaupt – zu verlassen, Haasau zu müssen.
Visitors of websites should yield: Besucher von Webseiten (not Besucher der Webseiten)
Note that translator-widget is rendered three different ways: As translator-widget , as Übersetzer Widget (good), and Konvertierung Widget (conversion widget).
… that have the Microsoft Translator Widget embedded should be: die das Microsoft Übersetzer Widget eingebettet haben.
Note: It would probably help the translation engine to put a clause like this inside commas.
can translate the content on the website should result in:
können den Inhalt der Webseite übersetzen – Note visitors – Besucher is plural so use können not kann, but content – Inhalt is singular.
without having to leave the site at all is rendered well except for the extraneous hyphen and the embedded Haasau (no idea how that got in there). … ohne die Website überhaupt zu verlassen zu müssen.
Awesome is one of those American English idiomatic expressions that rarely is translated correctly. I don’t envy you having to teach the engine equivalent expressions. Here is one: – which is awesome – was großartig ist.
I won’t go on – see, we just can’t get rid of contractions – but if my comments are helpful, I’d be happy to help more intensively.
I really like how Microsoft is going with the Microsoft Translator. It's also reassuring to see that someone from the team, like Vikram, really is paying attention.
I hope to see MST really get the syntax, grammar, and whatever issues aside to produce a machine translator that is almost as fluent as a real translator someday.
alexfung: Great comments. I have passed them along to our linguist/data teams. We constantly train the system on new data and do automatic and human evaluations to deliver comparable/better quality than other MT engines. As you noted, translation is a tremendously difficult job - particularly when you are building an engine that is targeting the general domain (it's easier if you target a specialized domain like medicine)
This is feedback that we have received from multiple folks and is high on our feature list. For now, you can return to the original page by closing the translation toolbar that appears during translation.
Thanks for the great feedback!
Microsoft Translator Team
Yes, the widget works fine. I test it from English to Spanish. Of course, there are few words that can be changed for a better understanding. To me, is a 90% understandable
I concur with the above poster, seems that germanic languages are notoriously hard to do by machine translations.
for instance: the Blogpost title reads:
Windows Experience Blog - Uit de Microsoft NAT pictogram inschakelt testen.
Normal Dutch version :
Test de Microsoft Vertaler 'widget'
Grammatically, we're all over the place but nowhere near what it should read like.
I've tested the translator translating the blog post into german. And the result isn't very good, hardly readable. Often the syntax isn't correct. I have the feeling the translator just translate word for word which is unfortunately often insufficient.
One more thing I noticed: The translator does not understand things like this: "I'd", it doesn't detect that theres a verb and just tanslate it to "Ich'd".
Well it works great, and the translations of this page are very fast.
The only thing I do not like about the widget is that everything is not controlled in the widget box, for example when you translate a page from English to Spanish, there is no option within the widget to return i.e. translate English to English by providing English on the drop down menu, or a button to turn off the translation on the widget.
Its easy enough to click the bar at the top of the screen to turn off a translation, but it would be nice to have a return to native language button on the widget, a little back arrow or something would suffice, that way I dont have to move my mouse very far.
Microsoft's Translation technology rocks! Can't wait for more languages to be supported (especially Indic). Also, I'm surprised we don't yet have a Sidebar gadget for Live Translator although there's a Live Toolbar button. Gadget please?
Try translate this page to Japanese. Whenever the text says "Microsoft Translator Widget", the translated Japanese text says "Microsoft Account-Alerts-Barcode-Calendar-Clipboard-Contacts-Core-Custom".
BTW, MSTW keeps the word "Widget" when translating to Chinese. Yahoo translated it to something probably correct but incomprehensible to me. Although Chinese is my mother tongue, I can only understand technical terms and jargons when they are presented in English.
alexfung, what did the Microsoft Translator Widget translate as "Microsoft Account-Alerts-Barcode-Calendar-Clipboard-Contacts-Core-Custom"?
Japanese. It is funny. You can look at it even if you don't understand Japanese. MSTW is translated as "Microsoft Account-Alerts-Barcode-Calendar-Clipboard-Contacts-Core-Custom"! Who can propose a theory why?
French. I haven't really read through the transated article. But it is strange that the first character after full stop isn't capitalised.
alexfung, thank you for your feedback! Much appreciated!
I read it in Traditional Chinese. It is generally readable but funny:
"Back during..." is translated literally. It becomes "The MSWT is now back from the process of MIX09". It is one of the places Yahoo did better.
"people with websites" became "people infected with websites" (like with a disease). yahoo does this phrase correctly.
"awesome". The translation means more like "awestruck" (as if it were a god). But Yahoo does exactly the same thing.
All the shortforms did not work: 've, 'd, 's. They are left as is, untranslated. Same problem with Yahoo.
"Below" becomes "noodles". In Simplified Chinese, they are the same word. In Traditional, they are different word with the same sound.
"if you are fluent...". All the parts of the sentence are translated reasonably, but the word order is completely messed up and it will take a bit time for a reader to figure out what you are saying. This also happens with the longer and more complicated sentenses.
"will be appreciated". Translation is literally accurate, but in Chinese we don't use the future tense in this situation. So it reads weird.
"on how well" becomes "how many". I can't understand how it can get wrong here.
"specially". Seems the translater is particular weak in handling adverbs at the end of the sentence. It just drop the translated adjectival form at the same location. I can't guess what it meant until I re-read the original English sentence.
"leaving that feedback on", translated in the sense of "keeping it there, leaving it behind". Yahoo on the other hand, treated "leaving" in the sense of "get away", just as amusing.
"register for an invite code" became "register to be an invite code".
In general, elementary sentence parts are reasonably translated with occasional intriguing mistakes. Longer sentences are usually messed up because the translated can't completely understand the structure, and even if it does, cannot easily map it into another language.
The overall quality is comparable to Yahoo. I think the MSTW handles some sentence fragments better: "was announced.", "- which is [awesome]", "I would love to", "add the [MSTW to you website]". It indicates that MSTW has hardcoded specific knowledge on these structures. On the other hand, you can see that MSTW did some funny stuff where Yahoo did much better.