| Pages in topic: < [1 2 3 4] > | Post-editing machine translations is a misnomer but there are now training sessions for it Thread poster: Bernhard Sulzer
|
|---|
Daryo
United Kingdom
Local time: 02:24
Serbian to English
+ ...
| Machine PSDEUDO/QUASI Translation | Feb 20, 2015 |
is the proper name.
Step one of the process of translation:
UNDERSTAND the ST
As of today, no one has managed to create a machine capable of that.
Only protein based devices are capable of that (albeit to a wildly varying degree).
ergo: "Machine Translation" is no more than marketing BS for something that should be quarantined in AI research laboratories for the foreseeable future, and only let out on a very short leash after passi... See more is the proper name.
Step one of the process of translation:
UNDERSTAND the ST
As of today, no one has managed to create a machine capable of that.
Only protein based devices are capable of that (albeit to a wildly varying degree).
ergo: "Machine Translation" is no more than marketing BS for something that should be quarantined in AI research laboratories for the foreseeable future, and only let out on a very short leash after passing the Turing test, not let loose on unsuspecting gullible public to turn them into paying guinea pigs for something that is still very far from being even a beta version.
Not to speak of the "verbal pollution" it's creating.
thus "Post-editing machine translations" is a non-existing job - how can you edit something that IS NOT a translation?
or to put it more bluntly - a con job on translators to make them do translations at lower price under the spurious pretext that the text was supposedly already translated.
[Edited at 2015-02-22 21:46 GMT] ▲ Collapse
| | | | Tom in London
United Kingdom
Local time: 02:24
Member (2008)
Italian to English
Fiona Peterson wrote:
London's population, is now 50% Muslim, I hear.
You heard wrong. Figures from the 2011 census show that the Muslim population in the whole of the UK was approximately 3 million. In large cities like London and Manchester they make up around 14% of the population.
[Edited at 2015-02-21 13:56 GMT]
| | | | | Totally agree... | Feb 21, 2015 |
Tom in London wrote: Fiona Peterson wrote:
London's population, is now 50% Muslim, I hear. You heard wrong. Figures from the 2011 census show that the Muslim population in the whole of the UK was approximately 3 million. In large cities like London and Manchester they make up around 14% of the population. I don't know where you heard different. In any case, whatever the figure, it isn't a problem; it's an asset. [Edited at 2015-02-21 13:15 GMT]
well said!
| | | | Jeff Whittaker 
United States
Local time: 21:24
Member (2002)
Spanish to English
+ ...
|
|
|
Bernhard Sulzer
United States
Local time: 21:24
English to German
+ ...
TOPIC STARTER | More, cheaper and faster | Feb 22, 2015 |
I didn't have time earlier to reply to all these points.
Let me say that I am always open to learn and good arguments can win the day. But I think it's good advice to everyone when I say be careful what you get yourself into when dealing with MT post-editing.
Read up on the reasons for its implementation (mostly - we want to pay as little as possible for as many words as possible simply because translations are so expensive) and the increased demand by outsourcers. There is an incre... See more I didn't have time earlier to reply to all these points.
Let me say that I am always open to learn and good arguments can win the day. But I think it's good advice to everyone when I say be careful what you get yourself into when dealing with MT post-editing.
Read up on the reasons for its implementation (mostly - we want to pay as little as possible for as many words as possible simply because translations are so expensive) and the increased demand by outsourcers. There is an increased demand to get more and more translated which would cost more of course the more words a client needs - thus it makes sense to look for ways to get more done cheaper and faster. Do more, cheaper and faster.
Samuel Murray wrote: Bernhard Sulzer wrote: >> quote from the training session info:
Special emphasis is placed on the differences between the skills required to revise the work of junior translators on the one hand, and those that are vital to correct MT output on the other, as this is the key to the successful adoption of MT in professional environments. Yes, the skills required (or: the techniques used) to edit a machine translation are completely different from the skills used to edit translations by inexperienced human translators. I see this confusion quite often, among both translators and clients, i.e. that editing MT is similar to editing a poor human translation, and it is not. Yes, both are "poor", but they require different skills or tactics to "fix".
By changing the MT output around (no matter how different this task is from normal editing) to achieve an acceptable translation, you're ending up with something else, most likely, than what you would have translated yourself from scratch. The task is very different from real translating and real editing because you will look at the source text and compare it with the MT output. You're not thinking "how would I now translate that" first by just looking at the source text. You will continue making changes of the MT output, based on the structure already given by the MT system output. I can see a tendency to edit in a way that will stay closer to what the machine produced than to what you would have translated yourself - something that will become routine the more you do this kind of work. But that's something for research studies. Does this possible change matter? And to whom? Will people care about how it's written as long as they can grasp the meaning? Could depend on the field.
If the MT output is excellent because the MT system is translating a text that is available in very similar form already and it is indeed a good rendering of the source text in general, you could indeed argue that you might be done faster with the translation and can do more in shorter time. But it's you that actually renders the "acceptable translation" - aren't you more than an editor?
Therein lie quite a few problems -
If MT would start to perform really well: Are you going to edit 10,000 words instead of translating 3,000 per day? Should/can you really read 10,000 words a day as some studies claim or suggest as quite feasible, in general?
Problem 1: let's say you have a complex text that needs a lot of revision because the MT sucks. But you agreed to a certain rate, let's say 60% of what you would charge for translating - you're exhausted after doing extensive editing of 5,000 words over 4 or 5 hours - which is really closer to re-translating than anything else - for a cheaper post-editing rate because it was supposed to increase your productivity?
Problem 2. What productivity is maximized here? If you end up spending close or a little more time on this job than it would have taken you to translate it yourself, you might have increased your daily output (5,000 words instead of 3,000) but I don't think you have increased your profit margin. And how good is the final text anyway if you could compare it with what you would have translated from scratch?
Problem 3: post-editing is certainly not about increasing your profit margin - it's existence has been justified by the need of outsourcers and end-clients to buy more words for less money - they simply need more words translated faster, thus they must cost less and MT is the way to do it, so we're told - and the training session will show you how.
Problem 4: the more people give in to lower rates for post-editing and taking on all kinds of MT outputs in various fields (don't get me started on technical manuals or medical/technical documents), we will not only hurt ourselves financially in the process (and thus respect for the profession will continue to diminish), but there are all kinds of other dangers lurking out there in terms of liabilities for the translations we finished which were started by a "machine" (MT system).
If you look closely, the whole issue is very complex and I refuse to simply accept that, generally speaking, I can be trained to maximize my productivity - and even if that's the case - to then charge less for what really are finalized translations, not revisions? Just because there is so much to translate and clients demand it must be done faster and cheaper, we have to use MT and charge less? Is that really all you need - someone to demand it to be done?
Samuel Murray wrote: Quote from the description of above post-editing training session: The focus is on the linguistic and translation-related phenomena that are particularly challenging for MT systems of different types (rule-based, statistical or hybrid). Yes, these different types of MT systems will make different types of "mistakes". Knowing what they are and how to fix them will make you more productive when dealing with these types of texts.
We should be more concerned about the basic quality of MT output - that's what would matter to me - and not accept various forms of bad/inadequate MT outputs and learn how to fix them - at various "fuzzy" editing rates (well let's analyze how many words you changed)? But always cheaper than translations, that's what is always implied by the people who demand these post-editing jobs and by researchers who point to low rates for such types of work as expected by customers and even by translators who readily participated in such studies - who are these translators anyway and what do they think of increased usage of technical assistance in translating and the oh so logical demand for ever cheaper rates? And who pays for/supports these studies?
Samuel Murray wrote:
By analogy, suppose you have to do a monolingual edit of two texts written by a non-native speaker of English (one written by an Italian person and one written by a Chinese person). If you know both Italian and Chinese, it would make it much easier to edit the two English texts than if you only knew that they were written by non-natives. And knowing that they were written by non-natives instead of e.g. school children will also help you do your editing more productively.
If you know Italian and Chinese. But that's still different from the situation where you have to go back to an OT text, then look at the MT output and then "edit" accordingly.
Samuel Murray wrote: Quote from the training session info: The differences between light and full PE are explained, in relation to different quality needs. I wonder... is this similar to light monolingual editing and full monolingual editing? I see many translators offer both services -- a light edit (in which spelling and major grammar issues are fixed) and a comprehensive edit (in which minor grammatical and also stylistic issues are fixed). And I know that some translators/editors are unable to distinguish between the two, so they end up doing either a comprehensive edit when it is not required, or a light edit that is too light.
That's some nice jargon - and where does the light monolingual edit end and the full (not just monolingual) edit start (well- by looking at the source text ) - and it's something that I really would reject: let people who ask for a light edit read the raw MT output and then let them read your light monolingual edit - how much are you going to change and what kind of final product are you supporting here? Did you look at the source text?
As a professional, I would never want to offer anything other than a good translation for an adequate price, not a bad translation or light monolingual edit for peanuts.
Samuel Murray wrote: also a quote from the training session info: You will learn how to incorporate PE into your translation workflow to maximize productivity. Yes, this is really something that must be learnt. Productivity isn't maximised automatically by simply including a technology into a process if you don't know how to do it right. Although I believe it is possible to maximise productivity with PEMT, I also suspect that in many cases where clients ask for PEMT the work ends up taking longer, and produces quality below the accepted standard, notably because the client incorporated PEMT into his workflow incorrectly.
The whole concept isn't acceptable to me - Are you participating in this training session?
Samuel Murray wrote:
I hope that when the trainer says "you will learn how to" he doesn't simply mean "you will learn that it can be done, in a few rare instances", without actually saying how.
It's either that or a demonstration with simple texts that a certain MT system has rendered on the basis of very similar texts it has processed before.
But be that as it may, whatever you call it, it should be integrated into the translation process, not replacing it. Me thinks.
The goal (of clients and outsourcers) seems to be to pay less and less because there are more and more words to be translated and people don't want or can't pay more and they don't extend the respect our profession deserves - does a Mercedes become cheaper because you need more and more Mercedes? Or what is it they want - clunkers? Very doubtful. Is it clunkers you want to produce? I don't.
==========================
Bernhard Sulzer wrote:
It seems a foregone conclusion that machine translation post-editing is a real theoretical concept with practical applications ( = makes sense in/for translation theory and practice).
Samuel Murray wrote:
The same can be said about "fuzzy matching", which works quite well in some languages and texts, and not very well in others. And it's a pity that CAT users receive little training (if any) in the difference between dealing with fuzzy matches, 100% matches and non-matches. It is just assumed, I think, that CAT users will "pick up the skill" to handle matching along the way.
It works very well for outsourcers/translators who apply an arbitrary discount scheme for fuzzy matches. It's another thing that's wrong in our industry. a) the more technology you use means you have to work for less money and b) that repetitions of words will determine the rate/fee for translation. When you use technology correctly, the knowledge of doing that should be part of the fee.
Again, to me it shows how little respect many translators have for their own work and where they're headed with such behavior.
BernhardSulzer wrote:
I am surprised about the degree of acceptance of this "concept" by Proz.com and the suggestion that (effective) post-editing of machine translation can maximize your productivity.
Samuel Murray wrote:
You can only be surprised if you have previously turned a blind eye to it. Or are you simply being hyperbolic when you use words like "surprised" and "alarmed"?
No, I meant it sincerely. Henry says this site was created for translators and that it has their interest in mind. If no one questions it anymore, what's going to happen? What has already happened?!
Bernhard Sulzer wrote:
Machine translation [is] a misnomer by the way because machines do not "translate" the specific text in context but compare words/phrases/sentences with similar phrases/sentences as translated before from anywhere really (unless a specific source is defined), based on algorithms...
Samuel Murray wrote:
I'm glad that you come out with this (i.e. your specific definition of "translation") right here in the first post of this thread, and not only in subsequent posts, as you have done in the past.
Now all we need to do is wait for the admission that the real reason you regard PEMT as a nonce is simply because of the "T" and the "E" in its name, i.e. that if it were called e.g. PFMM (post-fixing machine matching), then you would have had no problem with the concept and possibly even embrace it. Am I right?
Calling it what it is would help translators understand what it really is. But "post-fixing" still implies it's somehow a less arduous task or less work (and always so!) and cheaper rates are in order. Try to argue against that if people already tell you that your productivity is going to be maximized and you're going to do 10,000 words in the same time that you used to do 3,000 - for much less per word of course. And once the "machines" have gathered enough of your text to regurgitate it in future MT output, you'll increase your productivity again and again until you'll make less than what is considered minimum wage in many countries of the world. Well, as you said, the whole task is different from regular translation, so I hope the output will be bad enough to discourage translators from dealing with this while they get burned financially.
Again, we need a debate on principles before we jump into training sessions. But maybe some readers will report back here after the sessions. There's hardly much feedback available yet from our point of view. It's mostly from researchers and the producers of MT systems.
Bernhard Sulzer wrote:
Even though editing another person's translation can be called an editing task or a revision of a translation, it can only be called that because it was translated by a person who actually looked at the specific OT, the whole document, the whole context and thought about it and then proceeded to write...
Samuel Murray wrote:
I think you assume too much about what steps fellow-translators take during translation. (-:
And I do think that the machine looks at context... in an indirect way. A financial text will have different phrasing than a medical text (in both the source and target language), which means that the machine is likely to select the appropriate style anyway, because the style depends largely on the words that were used.
I wouldn't have any medical or technical text "translated" by a machine. There are many pitfalls. You'll see if you read the two articles I linked below.
But yes, I am not saying the various MT systems are completely useless. But we are sold down the river with this and the makers want us to believe we need to "replace" the human "translator" so we can be more productive (as post-editors) for our clients (more words in shorter time for less money) - that's what it comes down to.
Bernhard Sulzer wrote:
The exposure to incorrect language [such as many grammatical and stylistic errors and figurative meanings and concepts appearing as literal "translations"] is an utterly ridiculous basis for an "editing" job.
Samuel Murray wrote:
What do you mean by "exposure to"?
Having to read mumbo-jumbo. It messes with your brain and your nerves although you don't seem to think so.
==========================
Samuel Murray wrote:
To sum up:
The problem here is that you have a very specific meaning of the word "translation" and a very specific meaning of the word "editing" in mind that you expect everyone to apply everywhere (and that you expect, in fact, that everybody does already).
Step back and realise for a second that when PEMT'ers say "translation" they mean something else than you do, and when they say "editing" they something else than you do.
What they mean is "translation" supplied by a machine and then"edited" by a human for less money. But what is expected is a translation, let's not forget that.
Samuel Murray wrote:
In fact, PEMT doesn't even involve a machine, in the regular sense of the word "machine".
Machine in the general sense; software or platform specifically, or MT-system.
Bernhard Sulzer wrote:
... and the term "post-editing" has already been abused by outsourcers who use it to pay less for "post-editing machine translations" than for "translations."
Samuel Murray wrote:
Yes, and that is why modern translators need to receive additional training that deals with technology and processes that were not an issue in the pencil-and-typewriter age of translation.
Not what I think. I say it means we need to take a step back and look at this whole scheme and ask some very important questions before we jump headlong into something that sounds great but harbors many risks in terms of copyright issues, intellectual property, the future of our profession, the rate development, liability for mistakes caused by "machines" and then overlooked by humans (light editing?!), and many more.
Bernhard Sulzer wrote:
I don't see how a training headline like "Maximize Your Productivity with Effective Machine Translation Post-Editing" is helping this community of translators understand what this really implies...
Samuel Murray wrote:
It is an accepted custom in the Western media culture to use hyperbolic language for slogans, titles, and other interest-piqueing chunks of text. You must not interpret hyperboles literally.
Well, ask those outsourcers using the concept how they interpret it. There's nothing hyperbolic about the cheap rates they want to pay.
==========================
Bernhard Sulzer wrote:
I am surprised about the lack of interest in this topic.
Samuel Murray wrote:
You have begun several thread on ProZ.com about this very same topic (what is the neutral, non-offensive term for "troll"?), so perhaps the lack of interest is not in the topic but in the event (the "event" being: a dissing session about machine translation).
What ever made you decide to call me a "troll", even if you mean it in a neutral, non-offensive way?!
[Edited at 2015-02-22 07:47 GMT] ▲ Collapse
| | | | Bernhard Sulzer
United States
Local time: 21:24
English to German
+ ...
TOPIC STARTER | More general information on the topic of post-editing and MT | Feb 22, 2015 |
Here are two articles that are very good (IMO):
http://translationjournal.net/journal/29computers.htm
Quotes:
Cognitive Processes
To understand the essential principles underlying machine translation it is necessary to understand the functioning of the human brain. The first stage in human translation is complete comprehension of the source langu... See more Here are two articles that are very good (IMO):
http://translationjournal.net/journal/29computers.htm
Quotes:
Cognitive Processes
To understand the essential principles underlying machine translation it is necessary to understand the functioning of the human brain. The first stage in human translation is complete comprehension of the source language text. This comprehension operates on several levels:
Semantic level: understanding words out of context, as in a dictionary.
Syntactic level: understanding words in a sentence.
Pragmatic level: understanding words in situations and context.
Furthermore, there are at least five types of knowledge used in the translation process:
Knowledge of the source language, which allows us to understand the original text.
Knowledge of the target language, which makes it possible to produce a coherent text in that language.
Knowledge of equivalents between the source and target languages.
Knowledge of the subject field as well as general knowledge, both of which aid comprehension of the source language text.
Knowledge of socio-cultural aspects, that is, of the customs and conventions of the source and target cultures.
Given the complexity of the phenomena that underlie the work of a human translator, it would be absurd to claim that a machine could produce a target text of the same quality as that of a human being. However, it is clear that even a human translator is seldom capable of producing a polished translation at first attempt. In reality the translation process comprises two stages: first, the production of a rough text or preliminary version in the target language, in which most of the translation problems are solved but which is far from being perfect; and second, the revision stage, varying from merely re-reading the text while making minor adjustments to the implementation of radical changes. It could therefore be said that MT aims at performing the first stage of this process in an automatic way, so that the human translator can then proceed directly to the second, carrying out the meticulous and demanding task of revision. The problem is that the translator now faces a text that has not been translated by a human brain but by a machine, which changes the required approach because the errors are different. It becomes necessary to harmonize the machine version with human thought processes, judgements and experiences. Machine translation is thus both an aid and a trap for translators: an aid because it completes the first stage of translation; a trap because it is not always easy for the translator to keep the necessary critical distance from a text that, at least in a rudimentary way, is already translated, so that mistakes may go undetected. In no sense should a translation produced automatically be considered final, even if it appears on the surface to be coherent and correct. ...
...It has long been a subject of discussion whether machine translation and computer-assisted translation could convert translators into mere editors, making them less important than the computer programs. The fear of this happening has led to a certain rejection of the new technologies on the part of translators, not only because of a possible loss of work and professional prestige, but also because of concern about a decline in the quality of production. Some translators totally reject machine translation because they associate it with the point of view that translation is merely one more marketable product based on a calculation of investment versus profits. They define translation as an art that possesses its own aesthetic criteria that have nothing to do with profit and loss, but are rather related to creativity and the power of the imagination. This applies mostly, however, to specific kinds of translation, such as that of literary texts, where polysemy, connotation and style play a crucial role. It is clear that computers could not even begin to replace human translators with such texts. Even with other kinds of texts, our analysis of the roles and capabilities of both MT and CAT shows that neither is efficient and accurate enough to eliminate the necessity for human translators. In fact, so-called machine translation would be more accurately described as computer-assisted translation too. Translators should recognize and learn to exploit the potential of the new technologies to help them to be more rigorous, consistent and productive without feeling threatened.
http://www.linguist.com/blog/machine-translation/machine-translation-post-editing-and-human-translation-business-uses-and-pitfalls-2/
A human translator is generally consistent in strengths and weaknesses, but MT is random and that creates a danger. The reader should never assume that the types of errors are consistent or one good sentence near the beginning means you will find an overall good quality level. MT is a minefield of random errors, and the most dangerous ones are those that are hidden in statements that seem correct, until examination by a post-editor reveals that the correct meaning is exactly the opposite of what appears in the MT! This phenomenon is actually more common today than years ago when MT was based on a dictionary plus grammar rules instead of the modern statistical approach. The latter draws on a very large database of sentences and phrases and the computer seeks the best match to the input source text. But suppose the best fit contains a “not” that isn’t in the source? Unfortunately this is reality and the reader must always be vigilant. It is also the reason why a post-editor must have an excellent knowledge of the source languages and check every target segment against its corresponding source. Cleaning up the target text so it reads reasonably well without checking against the source is not true post-editing. Such practice is particularly misleading if the post-editor applies technical expertise to correcting target terminology yet leaves the wrong meaning intact because there was no check against the source. The most important quality of a translation is its faithfulness to the meaning of the source text, and this emphasis is particularly important for MT where meaning is so easily lost in a jumble of partially comprehensible sentences. By any definition, computers have not reached a capability that can challenge human intelligence. - See more at:
and a link to a previous forum post:
http://www.proz.com/forum/money_matters/215371-rates_for_post_editing_machine_translation_texts.html
[Edited at 2015-02-22 05:42 GMT] ▲ Collapse
| | | | Bernhard Sulzer
United States
Local time: 21:24
English to German
+ ...
TOPIC STARTER | How low do YOU go? | Feb 23, 2015 |
Rosanna Casamassima wrote:
PE a new way to get a good job paying the least, I agree with Bernard
That's exactly the danger here. Good job meaning an accurate "translation." Now everbody think about it - how low do YOU go?
| | | | Bernhard Sulzer
United States
Local time: 21:24
English to German
+ ...
TOPIC STARTER | Crazy segment | Feb 23, 2015 |
Kevin Fulton wrote:
Some clients only need a "gist" translation, for example in a legal discovery where there may be hundreds, perhaps thousands of documents that need to be reviewed for relevance. In this case MT products may need human review to make them intelligible.
It's not a segment of the market I give much thought to nowadays, although I have done "gist" translations in the past without MT. I was paid on an hourly basis for this.
[Edited at 2015-02-23 15:25 GMT]
Gisting legal texts? You mean to say you "reviewed" MT output without looking at the OT? For legal purposes? You were paid paid hourly to do what exactly? Make the most unintelligible parts (of text produced by a machine) more intelligible?
| | |
|
|
|
Kevin Fulton
United States
Local time: 21:24
German to English
| Didn't use MT | Feb 24, 2015 |
Bernhard Sulzer wrote: Kevin Fulton wrote:
Some clients only need a "gist" translation, for example in a legal discovery where there may be hundreds, perhaps thousands of documents that need to be reviewed for relevance. In this case MT products may need human review to make them intelligible.
It's not a segment of the market I give much thought to nowadays, although I have done "gist" translations in the past without MT. I was paid on an hourly basis for this.
[Edited at 2015-02-23 15:25 GMT] Gisting legal texts? You mean to say you "reviewed" MT output without looking at the OT? For legal purposes? You were paid paid hourly to do what exactly? Make the most unintelligible parts (of text produced by a machine) more intelligible?
Please read my posting more carefully. I didn't use MT.
You need to understand the legal discovery process. It is a review of potentially hundreds or even thousands of documents for relevance to a lawsuit. Basically the defendant (and plaintiff in a countersuit) turns over contents of the file cabinet containing documents pertaining to the case. Time is usually of the essence in these cases, as attorneys use the documents in preparing their case and directing questions in a deposition.
In one discovery I was involved in locally, there were numerous documents that were totally irrelevant to the case at hand, e.g. expense account vouchers, menus, conference itineraries, even an exchange of faxes relating to a misplaced umbrella (this was before email!), nothing relating to patent infringement, and certainly nothing worth using human translation in the vast majority of the pieces of paper I looked at. The law firm gave me a set of relevant dates, personal names and other data pertaining to the designation of the process in dispute. My job was to provide a one sentence description of the content of each document (called it "gist" translation if you will). Documents containing references to the data I was provided with got a longer summary. This latter material was then vetted by an attorney who then requested a full translation as needed. Of all the documents I looked at, less than 25% merited a full translation.
In another discovery about 20 years ago, about a dozen translators were at work translating well over a hundred documents each (if my load was an indication), relating to a dispute between two multinationals. If the agency involved had used machine translation to sort out the irrelevancies such as TV schedules, condolence letters, opera programs, etc. the process could have moved forward more quickly and cheaply, and attorneys could have requested full translations of genuinely pertinent documents. Instead they paid full price for thank-you notes, vacation schedules, etc. I honestly don't know what the attorneys did with all this material, but I am confident that they cared about throughput.
My point is that MT represents a certain market segment. In specific instances, it can be a useful tool in separating wheat from the chaff when there is a high volume of documents to be handled.
| | | | DLyons
Ireland
Local time: 02:24
Spanish to English
+ ...
| Craciunescu/Gerding-Salas/Stringer-O´Keeffe article | Feb 24, 2015 |
Bernhard Sulzer wrote: http://translationjournal.net/journal/29computers.htm To understand the essential principles underlying machine translation it is necessary to understand the functioning of the human brain. The first stage in human translation is complete comprehension of the source language text.
If you know what the conclusions are, it's usually possible to find premeses that will allow you to arrive at them.
Essay topic: "The Craciunescu/Gerding-Salas/Stringer-O´Keeffe assumptions are more plausible than Searle's strong AI hypothesis. Discuss."
| | | | Samuel Murray
Netherlands
Local time: 03:24
Member (2006)
English to Afrikaans
+ ...
.
[Edited at 2015-02-25 11:02 GMT]
| | | | Bernhard Sulzer
United States
Local time: 21:24
English to German
+ ...
TOPIC STARTER
Samuel Murray wrote: Bernhard Sulzer wrote:
What ever made you decide to call me a "troll", even if you mean it in a neutral, non-offensive way?! The bad troll posts threads about the same old topic over and over again, only to get a response, but the good troll posts them in an attempt to stimulate actual debate. It is only unfortunate that so little new learning comes out of every subsequent discussion.
I'm not a troll and I take offense to it.
http://www.thefreedictionary.com/troll
6. (Computer Science) computing (intr) to post deliberately inflammatory articles on an internet discussion board
| | |
|
|
|
Samuel Murray
Netherlands
Local time: 03:24
Member (2006)
English to Afrikaans
+ ...
.
[Edited at 2015-02-25 11:01 GMT]
| | | | Bernhard Sulzer
United States
Local time: 21:24
English to German
+ ...
TOPIC STARTER | Gisting etc. | Feb 24, 2015 |
Kevin Fulton wrote:
Please read my posting more carefully. I didn't use MT.
This is what you wrote:
Kevin Fulton wrote:
Some clients only need a "gist" translation, for example in a legal discovery where there may be hundreds, perhaps thousands of documents that need to be reviewed for relevance. In this case MT products may need human review to make them intelligible.
It's not a segment of the market I give much thought to nowadays, although I have done "gist" translations in the past without MT. I was paid on an hourly basis for this.
Which is it? With or without MT?
Well, yes, you make it clear, but the term "gisting" is very often tied to MT (see below).
Kevin Fulton wrote:
You need to understand the legal discovery process. ... If the agency involved had used machine translation to sort out the irrelevancies such as TV schedules, condolence letters, opera programs, etc. the process could have moved forward more quickly and cheaply, and attorneys could have requested full translations of genuinely pertinent documents. Instead they paid full price for thank-you notes, vacation schedules, etc. I honestly don't know what the attorneys did with all this material, but I am confident that they cared about throughput.
Why translate to sort out? A translator can look at the source text and decide. Why let a machine produce some unintelligible stuff and then run it through another machine for certain keywords? Is that what you're suggesting?
In any case, humans will have to look at something (source text or translation) to determine what is relevant and what is not.
I wouldn't trust that process to a machine or base that process simply on MT output.
Problem is that postediting MT is used to imply that these "translations" are workable drafts for some more "editing" to make them perfect.
By the way, the term "gisting" as in the following explanation isn't something I agree with either:
http://en.wikipedia.org/wiki/Postediting
The light/full classification, developed in the nineties when machine translation still came on a CD-ROM, may not suit advances in machine translation at the light postediting end either. For some language pairs and some tasks, particularly if the source has been pre-edited, raw machine output may be good enough for gisting purposes without requiring subsequent human intervention.
Kevin Fulton wrote
My point is that MT represents a certain market segment. In specific instances, it can be a useful tool in separating wheat from the chaff when there is a high volume of documents to be handled.
You haven't used it yourself and are supporting it?
[Edited at 2015-02-24 19:06 GMT]
| | | | Bernhard Sulzer
United States
Local time: 21:24
English to German
+ ...
TOPIC STARTER
delete
[Edited at 2015-02-25 01:53 GMT]
| | | | | Pages in topic: < [1 2 3 4] > | To report site rules violations or get help, contact a site moderator: You can also contact site staff by submitting a support request » Post-editing machine translations is a misnomer but there are now training sessions for it | 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 » |
| | 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 » |
|
| | | | X Sign in to your ProZ.com account... | | | | | |
Login to reply/comment 
Translation news
