Pages in topic:   < [1 2 3 4 5 6 7 8 9 10] >
What's your opinion on machine translation and quality?
Thread poster: Daniela Zambrini
Orrin Cummins
Orrin Cummins  Identity Verified
Japan
Local time: 00:31
Japanese to English
+ ...
We can't stand still Jun 5, 2014

Tom in London wrote:

In 100 or 1,000 years there is no machine or any other automatic way of understanding this type of text and transforming its nuances and allusions into comprehensible (in my case) English.

Alas, these days there is a naive "nineteenth-century" belief that technology, the new religion, will eventually be able to do what we know how to do already. Why so many people are wasting so much time trying to create machines that do what human beings already know how to do, is something that intrigues me.


I agree with you 100% that MT is very far from being usable in the real world of translation. It is highly probable that everyone reading this will all be dead and gone long before any tangible advances are made in MT systems that are available to the average consumer.

However, since humans first began crawling around this ball of mud we have always searched for ways to make technology improve our lives. If we didn't "waste time trying to create machines" to assist us, we wouldn't even be having this conversation because the Internet would never have been invented (nor computers, for that matter). I mean, who needs email? The old postal service worked fine. Humans carried the letters the whole way (still doing it, I hear) and though it took some time they eventually got there.

What about the automobile? Horses were great for thousands of years. Sure you have to feed them and they eventually get tired, but hey...the little buggers actually reproduce on their own (never seen a car do that yet).

I could go on but you see my point. The problem is not technology. The problem is the unscrupulous people who try to take that technology and use it to cheat other people out of money. We see threads about this almost every other day. PEMT is one of the biggest cons ever perpetrated in this industry, and I don't see that changing any time soon unless people wise up and stop accepting these scam jobs. But if researchers want to spend their time trying to improve MT technology itself, I say by all means let them do as they wish. I think we all know artificial intelligence is not going to make the huge advances required to actually threaten our livelihoods...at least not in our lifetimes.

Although, trying to predict what will be around 1000 years from now is a fool's errand. I doubt that anyone in the Middle Ages could have ever imagined that we would one day be landing spacecraft on distant moons or instantly communicating with people around the Earth. Not in their wildest dreams.


 
2nl (X)
2nl (X)  Identity Verified
Netherlands
Local time: 16:31
AutoType, AutoSuggestion, AutoComplete etc. Jun 7, 2014

Miguel Carmona wrote:

Thank you Michelle.

I have known that function for a long time, but I guess I knew it by another name, maybe autocomplete or something like that.


Actually that's not the same as a text expander (Typinator, Typeit4Me, TextExpander, MacroToolworks, PhraseExpress, Breevy etc.), where you have to define the triggering string (like 'btw' for 'By the way:').

AutoType, AutoSuggestion, AutoComplete, AutoWrite, Predictive Writing are all smart aids in CAT tools that automatically determine what you have typed before and offer you to reuse it. No need to manually define a triggering string.


 
RWS Community
RWS Community
United Kingdom
Local time: 16:31
English
The simple answer... Jun 21, 2014

Tom in London wrote:

Alas, these days there is a naive "nineteenth-century" belief that technology, the new religion, will eventually be able to do what we know how to do already. Why so many people are wasting so much time trying to create machines that do what human beings already know how to do, is something that intrigues me.


... to this question is that there is simply too much content that could be translated to make peoples lives easier if it were cheap and fast. So things like hotel reviews, Knowledgebase articles, blog posts, social media (twitter, facebook etc.)... and many more things like this that amount to huge numbers of words a day. It simply makes no sense to try and translate some of it properly, and it would be far too expensive and slow with Human Translation. So 100% Machine Translation in these circumstances might be good enough.

Of course there is content being produced that is not suitable for Machine Translation at all and the accuracy and quality that only a Human can produce is essential.

This distinction is very important, and even when a human translator is asked to post-edit machine translation, if there isn't a clear understanding of what quality is required then nobody will be happy. The translator will hate the work because they feel the need to correct everything, and the customer will be unhappy because it takes too long and potentially costs too much.

The threads in this forum always seem to refer to the quality being poor because the comparison is made with the excellent work a human translator can turn out. But sometimes it simply doesn't need to be that good, and correcting everything defeats the object of using it in the first place.

I think post-editing takes a willingness to work with Machine Translation in the first place, and an ability to hold back the urge to correct everything if it makes enough sense to be understood. And of course you need a client who also understands this and not one who is mistakenly using MT because they think a light editing will get them a top quality translation.

Regards

Paul


 
Bernhard Sulzer
Bernhard Sulzer  Identity Verified
United States
Local time: 11:31
English to German
+ ...
All that glitters is not gold Jun 22, 2014

SDL Support wrote:

Tom in London wrote:

Alas, these days there is a naive "nineteenth-century" belief that technology, the new religion, will eventually be able to do what we know how to do already. Why so many people are wasting so much time trying to create machines that do what human beings already know how to do, is something that intrigues me.


I want to add that the enthusiasm of inventing and trying to improve technologies can happen at the expense of some human (communication) quality - and I am not talking about CAT tools but MT. Young people trying to comprehend what it is their communication partner on the internet/Skype is saying to them through the MT-ed translation. This seems innocent enough but it has consequences built in - lack of communication in real social (if there is something like that) environments, exposure to a "word-mix" instead of accurate language - another topic. Often technology helps us and has more positive than negative aspects, granted.

SDL Support wrote:
The simple answer ... to this question is that there is simply too much content that could be translated to make people’s lives easier if it were cheap and fast. So things like hotel reviews, Knowledgebase articles, blog posts, social media (twitter, facebook etc.)...


Too much content that isn't worth being translated by humans?
Not sure if I want automatic machine translations of hotel reviews in 10 or 200 languages. Are you talking about statements typed by real people? Problem is as soon as sentences get a little complicated, machine translation will mess it up. If the review is a horribly written (in terms of language) machine translation, do I really want to visit the hotel? Well, maybe if it has 5 stars out of 5 next to it.

And what's the point of translating the reviews into many languages? We all know that English is quite a common language in the tourist industry - I would rather have translations of say a hotel review originally written in Chinese into accurate English and a few other languages maybe - and I would appreciate a few honest examples, not every single review.
How can I be sure I get the gist of the message a human wrote from MT alone?

Not sure how MT will help in social media - I can see us all succumb to trying to comprehend what the word-mix produced by a machine really means. I am not convinced I really need this anyway; again, a few main languages that people actually speak, even if they're not perfect in it - still better than a machine that never thinks or understands like a human - would be better IMO. Look at my comment here about a Skype conversation between two people with different native languages translated by a machine into their respective languages:
http://www.proz.com/forum/translation_news/269908-microsoft_unveils_real_time_audio_and_text_based_language_translation_via_skype.html#2303896

The conversation can be viewed here:
http://www.pcworld.com/article/2198140/skype-demos-will-launch-real-time-translation-app.html

Point here is that slurred speech, abbreviations, implied words and phrases etc. and many other things we do when we speak will not be understood by the machine; "understood' is the wrong word anyway, machines don't "understand" or "translate" like humans- but you alluded to that anyway by pointing to a need for quality translations done by humans.

Machine translation would work if enough phrases and sentences are stored in its memory so when someone says/writes something, the machine finds the same sentence in its memory. That's the only way it could work perfectly. Problem is, all future phrases and sentences have not been stored yet, and never will. Too many possible new combinations.

But you argue that less-than-perfect (I still call them bad) translations are okay as long as we can figure out what they mean. Again, I really don't like the fact that we associate the word "translation" with MT. To me, the act of translation is a human act.

Personally, I rather get things accomplished by communicating "correctly" and "accurately" in English or German. (or insert the languages you speak here).
Why should we be forced to communicate via "horrible" translations by machines?
Some may argue that's better than not communicate - I say communicate in the languages you and the person with whom you speak have a good command in. That's communication.
If you put MT between two people, it's indirect conversation, and it can be very difficult to get your points across.

Now, I am sure there are many applications where certain prerecorded words will be the only options available - meaning if your options in English are only "yes/no/don't know" - I am sure you could get that translated by a machine in as many languages as you like.

Your argument about cheap and fast - what is it really that we need to have translated so cheap and fast - I would appreciate a few examples. There's too much content? That's what some people say. There is a lot of content and probably more each year, but is it really so smart to have your city's website translated by a machine in all kinds of languages instead of having at least one "correct" English version of it?

Another thing - why should we acquiesce to the destruction (or dismantling) of correct language - that's what you get when you look at an MT-text - we can see something similar in abbreviations used with texting - R U ready? Granted, it's not a wrong use of the language yet because it's simply an abbreviation/letter symbols that is/are still understood as "Are You," but I see no advantage to readers and writers who will accept as an answer a sentence (if you can call it that) that is grammatically and stylistically atrocious. We should be more concerned about saving and/or expanding the sophisticated use of languages than to get used to picking meaning out of piles of garbled language. No thank you.
There's got to be a better/smarter way.

I am going to argue that if a company thinks it's too expensive to have a text (and I mean at least regular sentences, not just words with one specific meaning out of many meanings) translated into the 5 or how ever many languages they want by a machine to save money, it's probably wiser not to have it translated at all or have it done by a human, one language/market at a time. It will be better understood and appreciated by the target audience. We can discuss this a little more but, as a rule of thumb, I propose that "cheap" is never a good principle if you want to expand a business into new markets and it's bad in personal contacts because it displays a great ignorance and arrogance directed towards languages one doesn't speak or doesn't want to learn to speak or understand, and I don't think it's very respectful even though the people who are developing MT are probably proud of it.

That might sound a bit extreme and yes, buying a ticket shouldn't mean you have to learn a new language but there are ways to do it through people who speak your language(s) and other languages if these other languages are indeed needed for a particular reservation/purchase or other service. We're also talking about loss of jobs and of "human" interaction: here, again on a level where humans communicate through languages that they are capable of speaking correctly. If everyone just learns one of the main languages or the language of the country they now live in (if it's different from where they were born) as a second language, maybe the world would be a better place.

SDL Support wrote:
... and many more things like this that amount to huge numbers of words a day. It simply makes no sense to try and translate some of it properly, and it would be far too expensive and slow with Human Translation. So 100% Machine Translation in these circumstances might be good enough.


It makes no sense?
I don't agree. I believe that there are huge amounts of texts out there that should be properly translated - I wouldn't simply tell the search engine to use its built-in MT to give me the German version of an English website for example - those pages are horrible. Those are "cheap" translations "polluting" the internet. I would never read them.

I also believe there are texts that are voluminous - articles, blogs, etc. - but they don't change their texts from day to day, instead, new articles and blog entries are added every day - they could/should also be translated by a human translator if at all. I think human translation is warranted much more often than it's not. As far as expenses are concerned, it makes no sense to have something translated that no one can understand or you have to sit down and "figure out" what it means. Some online media have their articles translated into other languages by humans or have their associates create similar content in many languages - http://dw.de is a good example. Now that makes sense.

The main problem I'm having I guess is with the idea that "so much" needs to be translated as cheap and fast as possible - best by a machine that does it in "no" time. Who really wants these translations - texts that are really bad?

There are many texts that should be translated - but immediate, cheap, as a hardly comprehensible "word-mix" and by a machine don't strike me as important/good requirements - quite the opposite: within a reasonable time, at an adequate/fair price, accurately and by a human make much more sense to me in most circumstances; or better not translate them at all.

SDL Support wrote:
I think post-editing takes a willingness to work with Machine Translation in the first place, and an ability to hold back the urge to correct everything if it makes enough sense to be understood. And of course you need a client who also understands this and not one who is mistakenly using MT because they think a light editing will get them a top quality translation.


I wrote here - http://www.proz.com/forum/money_matters/270847-machine_translations_postediting-page2.html#2312282 - about how I dislike the term "post-editing" MT because it implies we are dealing with a real translation that simply needs some editing - when it is not a translation at all, in the "human" sense of the word. The task is not post-editing; it is "translating the original text while comparing/looking at an MT-ed version of it." Some say you don't have to type so much. It takes a while if you have to change every single sentence - rearrange words, replace wrong words with correct ones, fix punctuation mistakes etc.
So, no, I am not willing to post-edit MT. I am willing and appreciate the help of technology in my work - and your company is no exception - but not as a replacement for the human task of translating that needs a little post-editing by a human.

You write "I think post-editing takes ... an ability to hold back the urge to correct everything if it makes enough sense to be understood." Question is where do you start and where do you stop with this task? Also, what might be enough for one person to understand might not be enough for another. In any case. you are suggesting that grammar and sentence structure won't matter so much as long as one can look at the words and make sense of them. My experience is that MT doesn't give much text from which the correct meaning (as conveyed in the original text) can be deduced. Part of the problem is that MT can very easily grab a lot of "wrong words" and completely ruin the original meaning and render the text meaningless or incomprehensible. Maybe not always, but many times, especially when sentences are longer and more complex. I believe the strength of MT to pick the correct words or a mix of words that allows to grasp the gist of the intended meaning as conveyed in the OT is very overrated.
But there is also something simply wrong with having to look at some text and trying to get the gist, especially if the language is already in a horrendous and grammatically, stylistically and idiomatically horrendous state.
Maybe I am exaggerating but this is a serious topic about which we human translators should learn as much as possible = we shouldn't just nod our heads when someone says there are huge amounts of text out there not suited for human translation that can only be handled my machines for reasons of speed and money. It can easily give way to a lot of misconceptions about how important human translations are and can make certain people think they don't need us simply as post-editors of MT. Think about it.

MT - sounds good, is often presented as a great tool, but so far, it's really not - IMO. I don't know any convincing examples.
And we do need to distinguish MT - a machine giving you its version of a text in another language - from other technological tools (CAT tools, voice recognition) which are different from MT.

Here's a short clip on the difference between human and machine translation:
https://www.youtube.com/watch?v=_Ic_gph-2hU

By the way, Paul, I appreciate all your help with SDL products!

Just my thoughts.

Bernhard

[Edited at 2014-06-22 07:53 GMT]


 
Miguel Carmona
Miguel Carmona  Identity Verified
United States
Local time: 08:31
English to Spanish
... Jun 22, 2014

2nl wrote:

Miguel Carmona wrote:

Thank you Michelle.

I have known that function for a long time, but I guess I knew it by another name, maybe autocomplete or something like that.


Actually that's not the same as a text expander (Typinator, Typeit4Me, TextExpander, MacroToolworks, PhraseExpress, Breevy etc.), where you have to define the triggering string (like 'btw' for 'By the way:').

AutoType, AutoSuggestion, AutoComplete, AutoWrite, Predictive Writing are all smart aids in CAT tools that automatically determine what you have typed before and offer you to reuse it. No need to manually define a triggering string.


Got it. Thanks 2nl!


 
Samuel Murray
Samuel Murray  Identity Verified
Netherlands
Local time: 16:31
Member (2006)
English to Afrikaans
+ ...
What autocomplete is Jun 22, 2014

2nl wrote:
Actually that's not the same as a text expander (Typinator, Typeit4Me, TextExpander, MacroToolworks, PhraseExpress, Breevy etc.), where you have to define the triggering string (like 'btw' for 'By the way:').

AutoType, AutoSuggestion, AutoComplete, AutoWrite, Predictive Writing are all smart aids in CAT tools that automatically determine what you have typed before and offer you to reuse it. No need to manually define a triggering string.


One can be terribly semantic about this, but I agree with the previous poster, for I have known the text expander functionality by the name "autocompleter" for quite a while now. I realise, however, that several modern CAT tools have a function which they call "autocomplete" (or similar) which can automatically or semi-automatically suggest certain things.

You seem to think that the autocomplete function in all CAT tools should always work automatically, but nothing stops a CAT tool developer from combining what you call "autocomplete" with what you call "text expanding". OmegaT does it, for example. MS Word (not a CAT tool) does it too, but MS Word calls it "autocorrect".

OpenOffice (and some other programs too) also has a feature called "autocomplete", in which it tries to guess what word you're typing and then completing it. In fact, I was quite disappointed after my initial elation when WFC implemented "autocomplete" and I had to discover that it wasn't *that* kind of autocomplete that was implemented but something else which Trados users think of when they hear the word.


 
Michael Beijer
Michael Beijer  Identity Verified
United Kingdom
Local time: 15:31
Member (2009)
Dutch to English
+ ...
Clarification of terms being used Jun 22, 2014

Samuel Murray wrote:

2nl wrote:
Actually that's not the same as a text expander (Typinator, Typeit4Me, TextExpander, MacroToolworks, PhraseExpress, Breevy etc.), where you have to define the triggering string (like 'btw' for 'By the way:').

AutoType, AutoSuggestion, AutoComplete, AutoWrite, Predictive Writing are all smart aids in CAT tools that automatically determine what you have typed before and offer you to reuse it. No need to manually define a triggering string.


One can be terribly semantic about this, but I agree with the previous poster, for I have known the text expander functionality by the name "autocompleter" for quite a while now. I realise, however, that several modern CAT tools have a function which they call "autocomplete" (or similar) which can automatically or semi-automatically suggest certain things.

You seem to think that the autocomplete function in all CAT tools should always work automatically, but nothing stops a CAT tool developer from combining what you call "autocomplete" with what you call "text expanding". OmegaT does it, for example. MS Word (not a CAT tool) does it too, but MS Word calls it "autocorrect".

OpenOffice (and some other programs too) also has a feature called "autocomplete", in which it tries to guess what word you're typing and then completing it. In fact, I was quite disappointed after my initial elation when WFC implemented "autocomplete" and I had to discover that it wasn't *that* kind of autocomplete that was implemented but something else which Trados users think of when they hear the word.


To be honest, I'm now completely puzzled by everyone’s uses and different definitions of ‘autocomplete’ and ‘text expanding’.

In my opinion:

A. ‘autocomplete’ (feature in CAT tools that goes by many names: AutoType, AutoSuggestion, AutoComplete, AutoWrite, The Muse, Predictive Writing, etc.) =
CAT tool learns from your past typing (and connected TMs and TBs) and suggests words or longer phrases as you type.

B. ‘autocorrect’ (like in MS Word + Hunspell) =
words are corrected as you type, based on a list of correctly spelled words (such as the Hunspell dictionaries)

C. ‘text expanding’ (Typinator, Typeit4Me, TextExpander, PhraseExpress, etc.) =
series of letters can be defined to trigger/print words or longer phrases.

Incidentally, CafeTran has something called ‘Text shortcuts’, which is basically a ‘text expander’ (C. above).
It can be accessed via: Edit > Add selection to text shortcuts or pressing Ctrl + Shift + A after selecting something in the target box, and entering a trigger phrase or abbreviation (one will be suggested). ‘Text shortcuts’ you add while translating are added for the current session only, unless you do: Edit > List text shortcuts and click on ‘Save’. This will save them to a text file on your computer (@ ...\CafeTran\cafetran\resources\shortcuts\) called ‘shortcuts.txt’.

See:

Text shortcut
Tired of typing ‘hot air balloon’? The next time you type the word in CafeTran, select the term and click ‘Edit > Add selection to text shortcuts’ (Ctrl+Shift+A on Windows). The text shortcut dialogue will appear where you can define a text shortcut for the term (e.g., in this case, CafeTran will suggest ‘hab’). From now on, all you need to do is type this shortcut, and CafeTran will automatically insert the corresponding complete term.
(http://cafetranhelp.com/definitions )

See also the amazing (and ever evolving) AutoHotkey script ‘AutoCorrect’: http://multifarious.filkin.com/2013/05/31/autocorrect-for-everything/

Michael

[Edited at 2014-06-22 11:34 GMT]


 
Samuel Murray
Samuel Murray  Identity Verified
Netherlands
Local time: 16:31
Member (2006)
English to Afrikaans
+ ...
Definitions, definitions (as long as we all know roughly what we mean) Jun 22, 2014

Michael Beijer wrote:
To be honest, I'm now completely puzzled by everyone’s uses and different definitions of ‘autocomplete’ and ‘text expanding’. ... In my opinion:


https://en.wikipedia.org/wiki/Autocomplete
https://en.wikipedia.org/wiki/Autocorrect

In short, autocomplete (in the world out there) is when you type the first few letters of a word, phrase or sentence, which the program then tries to expand for you, and autocorrect (in the world out there) is when you type a sequence of characters that do not form the exact first few letters of a word, phrase or sentence, which the program then tries to expand for you.

Where the program gets its prediction information from (e.g. from previous words typed by the user, from a TM, from another part of the file e.g. the source segment, or from a list that the user created himself) is not really relevant... in the world out there.


 
Samuel Murray
Samuel Murray  Identity Verified
Netherlands
Local time: 16:31
Member (2006)
English to Afrikaans
+ ...
The world is catching up (and so are clients) Jun 22, 2014

Bernhard Sulzer wrote:
Not sure if I want automatic machine translations of hotel reviews in 10 or 200 languages. Are you talking about statements typed by real people? Problem is as soon as sentences get a little complicated, machine translation will mess it up.


Yes, but the world is apparently already used to that. I recently had a client ask me to translate a series of such reviews in such a way that it appears as if it is a machine translation, while making sure that the exact message that the client wants to convey with the reviews is not lost.

I wasn't quite sure how close to the edge of ethics it is to translate such reviews (they are fake reviews, after all, which the client will post under fake names, for his own product) and in particular to translate them in a way that deliberately deceives the user into thinking that they are not faked because they don't appear polished.


 
Michelle Kusuda
Michelle Kusuda  Identity Verified
United States
Local time: 11:31
English to Spanish
+ ...
It comes down to client-centered education! Jun 23, 2014

Samuel Murray wrote:

Bernhard Sulzer wrote:
Not sure if I want automatic machine translations of hotel reviews in 10 or 200 languages. Are you talking about statements typed by real people? Problem is as soon as sentences get a little complicated, machine translation will mess it up.


Yes, but the world is apparently already used to that. I recently had a client ask me to translate a series of such reviews in such a way that it appears as if it is a machine translation, while making sure that the exact message that the client wants to convey with the reviews is not lost.

I wasn't quite sure how close to the edge of ethics it is to translate such reviews (they are fake reviews, after all, which the client will post under fake names, for his own product) and in particular to translate them in a way that deliberately deceives the user into thinking that they are not faked because they don't appear polished.



Back in 1993 (yes, I am ancient!) a company which owned an automatic translation software program would hire me to translate the marketing text that appeared on their software boxes. Of course, it created the illusion that the translation program was better than it truly was.

I personally believed that the program had true potential. However, as operating systems evolved, compatibility issues emerged and they had trouble keeping up with Microsoft Operating Systems upgrades. The company was sold to another MT vendor which instead of upgrading the program, started taking features away, making it more affordable but rendering their program useless.

I am not a computer scientist so I do not know if even companies like SDL will be able to keep up with all the rapid changes!


 
Chié_JP
Chié_JP
Japan
Local time: 00:31
Member (2013)
English to Japanese
+ ...
Just improve TM usage before boasting your MT Jun 25, 2014

This is my second post and surprised there are dozens of people here.

Before they go boasting that engineers can kill off translators by MT if they want to and "cut human cost", they just have to improve your usage or TM functions.

Currently when your TM has a segment that says "turn to left",
and if the next translation is "turn to right" TMs I know of, at least Kilgray or SDL, do not even take care of the simple translation of "left" and "right",

... See more
This is my second post and surprised there are dozens of people here.

Before they go boasting that engineers can kill off translators by MT if they want to and "cut human cost", they just have to improve your usage or TM functions.

Currently when your TM has a segment that says "turn to left",
and if the next translation is "turn to right" TMs I know of, at least Kilgray or SDL, do not even take care of the simple translation of "left" and "right",

then asks you further to set parenthesis and tags around left-right, manually.
This is very tedious and simple work that I sincerely hope computers can take care of instead of language learned workers, but today TMs or MTs do not even know how to do this and say "costly, etc..." If they think they can cut cost drastically by replacing humans with computers, this is not the word we expected to hear because this indicates MTs are also too high to afford,
this means TMs and MTs are simply more costly than humans and they need to cut their computer and engineering cost first in order to replace costly humans to an extent that engineers get starved.

So far, the situation is out of question and before talking about MT,
I just ask them to talk about improving usage of TM and
I guess I already know the answer from TM agencies to ask us to pay 800,000USD to purchase a software license with auto-translation. Everyone quits and companies purchase them, and then outsource the rather complicated review task to part time workers with insufficient knowledge or training. If you read the almost illegible "Microsoft Machine Translated Help -waiver, this document may contain error",
it will be easy to understand how the documentation and translation is going to be like
with state of the art Machine Translation.
Collapse


 
Post removed: This post was hidden by a moderator or staff member for the following reason: empty post
Neil Coffey
Neil Coffey  Identity Verified
United Kingdom
Local time: 15:31
French to English
+ ...
Under a very specific set of circumstances Jun 25, 2014

Daniela Zambrini wrote:
Do you use machine translation? If so, how? If not, why not?


As a translator, I tend to look at things like this:

- the client wants a particular translation to a particular standard (or has some need with regard to a document in a foreign language) by a particular date
- in order to achieve what the client requires for a given project, is it more beneficial to have MT as "part of the mix" versus other alternatives?

When time or budget do not permit a "normal" translation, the "other alternatives" may include splitting the work among several colleagues, one colleague providing a "gist" translation, a conference call with the client where they ask questions about the content of the document, etc.... and... possibly a solution involving machine translation.

In other words, for me as a translator the use of MT is one potential part of the solution that is on the table to meet specific project needs. I have no ideological reason not to use it myself in principle if it turns out to be the right tool for the job.

However, in practice, I have found it extremely rare for MT to be the right tool for the job. I have used it a couple of times, for example, where a client has a very large volume of material, has a particularly high quality/accuracy tolerance and where the nature of the text (e.g. significant repetitions) means that MT combined with some pre-processing is a viable part of the solution. But as you can see, these are quite specific circumstances. Overall, my answer to the question becomes "practically never".


 
Neil Coffey
Neil Coffey  Identity Verified
United Kingdom
Local time: 15:31
French to English
+ ...
And... Jun 25, 2014

In my previous post I mentioned that "I myself as a translator" have no ideological reason not to use translation.

I do think that there is a separate problem engendered by the use of MT in that as soon as MT is involved, clients overestimate the capabilities of the actual MT component itself. Because a *combination* of (custom pre-processing, MT, post-editing) worked, in the client's mind, that means that "MT" is viable...


 
RWS Community
RWS Community
United Kingdom
Local time: 16:31
English
Too late now ;-) Jul 1, 2014

Bernhard Sulzer wrote:

By the way, Paul, I appreciate all your help with SDL products!



More seriously though Bernhard, I'm not going to try and answer every point you made here because in general terms I think you have missed the point I was making altogether. The question was why would people want to waste time creating machines to do things that people can do anyway... or this was the essence of the question I was answering.

I use MT on a daily basis. I use it to translate tweets from Chinese or Russian so I can make enough sense of it to point someone in the right direction; I use it when I'm looking for a hotel, or buying a product, to read the reviews (I'm not interested in the "glossy" advertising that's already been translated); I use it to translate a manual for some foreign goods I own where the manufacturer didn't deliver it in English; I use it when I purchase an incredible amount of stuff in Germany because they seem to provide stuff in German, Italian and Polish... but not English! I could probably go on, and I don't think I'm that different to many others either.

In an ideal world all of this stuff would be handed over to human translators who would be paid a good sum to deliver this material in every language available for anyone who cannot read the material it was originally delivered in.

But if I take as an example TripAdvisor. They run approximately two billion words a month through SDL BeGlobal. The scale of this is quite hard to grasp, and it's not unique to TripAdvisor. There is content being produced all over the place in multiple languages that people may want to read, but it would be far too expensive, and far too slow, to do it manually with human translators. This type of material is also not important enough to warrant a perfect translation, but deemed too important to ignore.

In answering this specific question this is where I am coming from. The scales we are talking about here are completely different and are only possible if people spend their time developing this kind of technology because people cannot do this anyway.

Regards

Paul


 
Pages in topic:   < [1 2 3 4 5 6 7 8 9 10] >


To report site rules violations or get help, contact a site moderator:


You can also contact site staff by submitting a support request »

What's your opinion on machine translation and quality?






Protemos translation business management system
Create your account in minutes, and start working! 3-month trial for agencies, and free for freelancers!

The system lets you keep client/vendor database, with contacts and rates, manage projects and assign jobs to vendors, issue invoices, track payments, store and manage project files, generate business reports on turnover profit per client/manager etc.

More info »
TM-Town
Manage your TMs and Terms ... and boost your translation business

Are you ready for something fresh in the industry? TM-Town is a unique new site for you -- the freelance translator -- to store, manage and share translation memories (TMs) and glossaries...and potentially meet new clients on the basis of your prior work.

More info »