Foreign (non English speaking) Members

[Matt Davis]

Originally Posted by Marty Welk

If you write simply, it can be translated.

Once, a judge of Court of Session of Scotland sent the Editor his candidate which reads: "In the Nuts (unground), (other than ground nuts) Order, the expression nuts shall have reference to such nuts, other than ground nuts, as would but for this amending Order not qualify as nuts (unground) (other than ground nuts) by reason of their being nuts (unground)".

I also refer the honourable gent, and to all herein, to the case of the most unsuccessful phrasebook, now known as 'English as she is spoke':

English As She Is Spoke - Wikipedia, the free encyclopedia

Alas, the Wikipedia entry does not mention the howlers, which you can find easily on the web. But thanks to the idiom that says 'to err is human, to really foul up requires a computer' (and to get us back OT) here's BabelFish v. English As She Is Spoke:

English As She Is Spoke vs. Babelfish!

[Paul Mailath]

And the Australian Version of that wonderful book:

Lets Stalk Strine by Afferbeck Lauder

http://en.wikipedia.org/wiki/Afferbeck_Lauder

You'll be speaking like a drover in no time

[Andris Krastins]

I can attest that the Google translator from English to Latvian is totally unusable.
But I hear it's a lot better with major languages. Any way, I know English well enough. :)

[Hans Ledel]

I just read the English to Swedish translation and I´m sorry to say it does no work at all

Cheers

Hans

[Peter Berggren]

I think most Swedish people between the age of 15 and 65 easier understand the English text than the machine translated to “Swedish”. The “Swedish” version didn’t make any sense.

Best regards, Peter

[Alex Khachatryan]

Originally Posted by Chris Soucy

Russian:

Иностранные (Non Английский говоря) Члены или возможных members.As простоя, Новый год День мысли (не то чтобы я был, другой тяжелый день в офисе, ER, работая в саду, и, пожалуйста, говорить, если у вас есть здесь раньше меня) мне пришло в голову, сколько именно, не английский (в качестве первого языка), выступая Члены есть на DVinfo.From там я получил на интересно только, сколько членов DVinfo действительно имеет в каких странах по всей планете? Я знаю, ты до примерно 26000 нечетные (некоторые из них действительно очень странный. Что, Кто, я?), Но где они? Оттуда оно было небольшим, но пропустить и перейти к, гм, как представляется, будет большая расхождения наблюдаются в плакатах между английской речи и без Английский выступивший частях планеты (Вы не говорите!). Теперь, учитывая оглушительный успех в функцию поиска Google добавил некоторое время назад (блестящая идея, почему я не думаю, его себе?), а в качестве помощи по март DVinfo завладеть известной Вселенной на тему, ER, DV, я подумал, может быть DVinfo браке с системой перевода Googles и стал почти универсальным? You'all должны быть знакомы с это: Язык ToolsThe вопрос, можно ли в браке с IP адресов для автоматического перевода сообщений / поиска для стран назначения, если необходимо или выключены для комфортного ESL * Член? Я не думаю, что вам нужно слишком беспокоиться о Элмер фадд (что может вырасти на вас Tho '), "Пираты, BORK и т.д., Hacker или клингонский переводов, особенно последний, Tho' Вы никогда не знаете, есть какие-то странные существа скрываются на эту site.The остальная часть (довольно много, на самом деле ) вполне может открыть новую дверь для DVinfo.Nope, я понятия не имею, насколько успешно система является, как только я говорю Языки являются Strine, киви, канадец и ПОМ, все из которых, как представляется, таинственным отсутствуют в списке - Хмм , есть теории заговора похоронены где-то там. И нет, это не двери назад DIG о региональных кодов (интересно тем, что он является) просто полный "думать" из коробки "момент (и первый, кто пришел с" New Years Day Out = деревьев, Не сейф, будет не слишком далеко .... ER, гм!). имеющими хорошую, Крис Г. и все, что вам DVinfo'ers. КСО: Это не drill.PPS: * = ESL Английский как Второй язык Последняя редакция Крис Soucy; 1 января 2010 в 06:58 PM. Причина: Whoops.

...

Anyone want to comment on the accuracy?

This is getting a lot of hits, but very little input, anyone want to chime in?


CS

There are just bunch of words but no meaning.
Note: I am native Russian speaker.

[Chris Soucy]

Wow, 1,644 hits as of this post, I would never have thought such an esoteric subject could or would generate such a level of interest.

I am pretty well convinced the "system" as it exists is entirely unworkable, thanks for the further confirmation.

Quite how the various messageing networks think they can use it or something similar has me completely baffled.

Still waiting to see if there's any feedback from "up there" on the subject.


CS

[Alex Khachatryan]

Chris, I do not want to act too critical, there are some parts when I read, I catch a slight meaning (3-4 words getting connected) but then, boom, it takes me nowhere again.
Over all, there is an improvement in results comparing with the ones I got 5 years ago, trying out automatic translation services.

[Chris Soucy]

Still got interest, but no feedback from the hierachy, guess it's not an issue they wish to persue.

Just to keep the ball rolling tho', cop this titbit of news, well, speculation, if their current offerings are anything to go by........

A phone that translates 6,000 languages in real time, really? - News, Gadgets & Tech - The Independent

Can't see it myself, but there you go.


CS

[Ervin Farkas]

English is not going to be replaced any time soon by any other language, especially when it comes to technology. If anything, the position of English is going to get stronger and stronger as more and more people are involved with technology.

I come from a bi-lingual background - born Hungarian in Romania, I speak both languages at mother's tongue level... plus try my best in English. Recently I helped a relative in Budapest put her computer in order software-wise via remote access. I can't tell you how much frustration it was to try guessing what the different Hungarian terms might mean in English.

So this is what I say: we better help everyone learn English instead of translating.

[Matt Davis]

Originally Posted by Ervin Farkas

we better help everyone learn English instead of translating.

English is such a silly, stupid, messy chopped up language full of arcane rules and spellings it makes Japanese look easy. Okay, so it has borrowed words from loads of languages, but there are so many subtleties and nuances that can trip people up. And have you ever experienced the tirade of abuse one can receive with a misplaced (aka 'Grocer's') apostrophe? I'm guilty of it myself.

If only it were Dutch that made it as 'the' lingua franca. Or something simple. I'd have been in favour of Esperanto, but always sounded a little odd to my ears...

But I digress. Have been filming and editing a few movies on the topic of networking, and Cisco recently held their big Euro event for networking engineers. The biggest topic was video. Cisco want to 'pwn' video plumbing.

They are almost ready with server-side smarts that analyse incoming video, does machine transcription for later searching, and then providing machine translation for subtitles/closed caption. This is not an 'idea' or a 'prototype', its what Cisco needs to make happen in order to make sense of this strange datatype which has, in the past, been treated like a month old skunk corpse by networking engineers.

It feels like there's other technology being sat on by the likes of these large organisations that do improve on current examples of transcription and translation.

[Ervin Farkas]

Not surprisingly, the offer to translate comes from English speaking people (talking about this particular thread here), meaning, from what I've read here, from people who don't speak another language. You guys need to understand that there is English, and then there is specialty English. You can bring in the best human translator in the world, if he does not know "multimedia English", he will not be able to accurately translate. Factor in the hundreds of specialty terms used with cameras, NLEs, etc, etc...

At this point, machine translation barely makes it with general, every day language - let alone any specialties.

Translation is a FORM OF ART, not a skill one can easily learn, and pretty much imposible for machines at this point. Over the years I had the chance to listen to translators who were speaking both languages perfectly, yet their translation was horrible. A good translator is born, not educated (of course, skills may and need to be refined).

Machines will never "get there". Here's a poem by the greatest Hungarian poet Petőfi Sándor, very simple words, nothing hard - translated to English by Google.

FÜSTBEMENT PLAN

All the way - home --
I'm thinking:
How do I call
Had not seen my mother?

What will I say first of all
Nice, nice to him?
When that rocking cradles,
The arm extends.

And I thought many
Szebbnél-better idea
While the time seems to stand,
Although the truck was running.

And the little room toppanék ...
Flies up to me with my mother ...
And I csüggtem lips ... silence ...
As the fruit of the tree.

Sounds horrible in English... doesn't it? Now think about what it would look like if you try to translate say embedding metadata or bitrate, color grading, white balance, or AVCHD...

Do you see where I'm going with this? Let's revisit this topic 50 years from today, we may have better options in terms of technology... until then, as I already said, let's just keep it good ole' English.

[Chris Soucy]

Franz Josef Och, Google's translation uber-scientist, talks about Google Translate | Technology | Los Angeles Times

and

http://www.nytimes.com/2010/03/09/te...translate.html


CS

[J. Stephen McDonald]

Here's a free translator that seems good to me-----but don't always trust the gender and singular/plural designations to be correct. These online translators work best if you have enough basic knowledge of a language, to notice the mistakes and be able to correct them. Many of them provide more complicated and often outdated phrasing, than would be used in contemporary conversations. Familiar forms are more commonly used these days with many languages, but the formal versions of verbs and pronouns are most often provided by the translators. When it comes to technical words or phrases, the English versions are often the ones learned and used by speakers of many other languages. In some cases, it might be better to use them in the translations, rather than a word or phrase from the other language, that may have been coined long before camcorders and computers were invented. This doesn't apply to French, of course. There's a government agency to keep that language free of anglicisms. Free Online Translator

[Jim Andrada]

Couldn't resist a couple of longish comments

1) A hundred years ago I got tangentially involved in machine translation while I was an undegraduate at Harvard in the Sputnik era (I also worked for the Air Force tracking Spunik 2, but that's a separate story) As you can imagine (or remember if you're old enoug) there was a great and sudden interest in translating Russian technology texts to English and the academic community wasn't shy about getting government research funding for the purpose and teaching courses in it. And I wasn't shy about taking the courses to satisfy some long forgotten course requirement.

The results at the time with the limited computer power we had in the 1950's were pitiful (along the lines of "The spirit is willing but the flesh is weak" coming out as the Russian equivalent of "wine good, meat rotten")

Disclosure - I don't speak Russian but then again most people working on the project didn't speak it either.

2) Fast forward to 1965 or 1966. I was working in an ad-tech group at IBM and we were trying to build a natural language query system. The impetus for the project, aside from looking for a way to sell more computers was the belief that people would be more capable of extracting information from data bases if they could query in natural language. I was working on the front end, ie the syntactic analyzer and intermediate code generator, which would have been followed by the query engine itself. I also got to run around the country presenting papers on our project at groups like the Association for Computational Linguistics.

It was all great fun and we were sort of able to satisfy a simple query in English, German, or Persian along the lines of "are there any books in our library about mathematics". Interestingly enough, the correct answer would have been "yes" or "no" but we took as an assumption that every query should be answered with some kind of list. I don't remember if we distinguished between books and papers and magazines, or not, but I have the funny feeling that we just lumped them all together.

While it might seem that we might have just been parsing the input for keywords (a la a Google search etc) we were actually doing a fairly sophisticated syntactical analysis of the input stream - which is why we kept getting invited to present papers I think. We also had a bunch of MIT and Harvard grad students in linguistics working with us as part timers.

Anyhow, our conclusion was that what people really wanted wasn't as much the ability to query in natural language as the ability for the computer to read their minds and figure out what they were looking for - they didn't want to take care to disambiguate or precisely state what it was that they wanted in the first place, and if they had to do it anyhow, they were just as happy with a structured way of representing the query.

Another conclusion was that real machine translation was probably an unattainable goal, largely due to (as someone said earlier) the need for disambiguation. And not just disambiguation in the syntactical sense, but also in the much more difficult semantic sense - ie understading not just the syntactic structure of the input stream, but also understanding what a word meant by itself and what it might also mean in the context of the overall discourse, slang, local dialect, social environment, etc etc etc. For example, how should we know that a nice kettle of fish would be one thing if you were in someone's kitchen and a totally different thing if the conversation were about someone who had gotten into a lot of trouble - ie was in a real jam, and not the kind of real jam that you spread on your bread or was maybe participating in a jam session. Or that the meaning of "fine point" would be different if you were discussing a legal issue or knife sharpening.

And this is all in the context of a single language - now exponentiate the difficulty by trying translation - how should the computer know that the appropriate translation of "Thank you" from English to Japanese is in some cases the equivalent of "Excuse me" or maybe silence?

(When someone holds a door or goes out of their way to do something for you it's considered socially appropriate to excuse yourself for causing them trouble - hence the response should be "sumimasen" and not "arigatoh" - unless of course the nice person was in a position where it was their job to open the door for you in which the right response would be to say nothing at all)

Google Translator as I understands it works on the basis of being a statistical translator - ie it searches for an example of the same or similar sentences or phrase in some already translated material - something like "I love you" should be duck soup as it's been translated from and to English and almost every language on Earth and the line appears in almost every romance that has been translated.

But what on earth would it make of "Crush the blacks"? Is this some Apartheid era racial epithet or maybe an admonition to a NZ football team or maybe, just maybe, some obscure techno-speak from a videographer? I ran a test from English to Japanese and the system did a great job of avoiding the issue by just replacing the English "crush" with the phonetic representation in Japanese Katakana and "Blacks" with the characters for "black person"

It also translates "Warm blacks" as meaning nice toasty black people. Haven't tried cool whites yet. By the way it gives differing translations for "Crush" or "crush"

This is very dangerous stuff, particularly if you don't know both source and target languages quite well - in which case why would you bother!

Disclosure: I'm a native English speaker and speak Japanese reasonably well, but not native speaker level, in spite of which it's been my home language for 20 years and a language I use for business nearly every day. Considering that I was almost 50 when I started studying it, I don't feel too bad about not being perfect.