It is neither new nor interesting to observe that the mention of machine translation (MT) provokes strong opinions in the language services industry. MT is one scapegoat for ever decreasing per-word rates, especially among independent translators. The choice to accept post-editing work is often cast in moral terms (peruse the ProZ forums sometime…). Even those who deliberately avoid MT can find it suddenly before them when unscrupulous clients hire “proof-readers” for MT output. And maybe you have had one of those annoying conversations with a new acquaintance who, upon learning your profession, says, “Oh! How useful. I use Google Translate all the time!”
But MT is a tool, and one that I think is both misunderstood and underutilized by some translators. It is best understood as generalized translation memory (TM), a technology that most translators find indispensable. More.
See: The Savvy Newcomer
Subscribe to the translation news daily digest here. See more translation news.
Spain
Local time: 07:54
Spanish to English
+ ...
By 'backoff', in the sentence "New translators should think about how to integrate MT into their workflows as a backoff"... don't they mean 'backup'? Or is this a new thing? Perhaps it's an abbreviation for 'backoffice tool?
[Edited at 2016-05-19 08:15 GMT]
[Edited at 2016-05-19 08:15 GMT]
Denmark
Local time: 07:54
Member (2003)
Danish to English
+ ...
Dinosaurs like me can simply opt out, but I have to admit I hear echos of my distant student days, when I trained as what was then called an Information Scientist. To avoid confusion, I prefer to call myself a technical librarian. Computers back then worked on punch cards, and had advanced to a whole line per card. There would be about one or two computers per university and very few in industry.
We really had a lec... See more
Dinosaurs like me can simply opt out, but I have to admit I hear echos of my distant student days, when I trained as what was then called an Information Scientist. To avoid confusion, I prefer to call myself a technical librarian. Computers back then worked on punch cards, and had advanced to a whole line per card. There would be about one or two computers per university and very few in industry.
We really had a lecturer who said that computers were fine for mathematical departments, but he simply could not conceive that computers would ever be useful in libraries. However, we needed to know about them so that we could keep them out!
Back then, before the days of the Internet, people used libraries, and we were the gremlins lurking about in the stacks (basements) who tried to find what you wanted if it was not on the shelves open to the public.
We learnt some COBOL and even wrote a program that could sort 100 titles alphabetically. THAT was a pain! Most of us would have written a list or laid the books out on the floor - and had them sorted manually before you could say Common Business-Oriented Language.
The point was that once the program worked with 100 titles, it would in principle work with a whole library... but boy, did we struggle with it.
Gremlins are smaller these days. Back then, so-called pocket calculators, like pocket dictionaries, weighed nearly a pound and were really only abacuses with pretty red lights when the battery was not flat.
Today, you can have the whole Internet plus a camera, a radio and a multitude of apps and gadgets on a thingy that really fits in your pocket. It even works as a phone...
There are still librarians at your university library, but they would have been put out to grass long ago if they had refused to work with computers.
As translators, we have to stop competing with computers, because they win hands down at some things. We have to find the things they can't do, and so far there is still plenty of work for humans. ▲ Collapse
United Kingdom
Local time: 06:54
Member (2008)
Italian to English
Dear Christine ,
A computer could not have had the thoughts you have, nor could it have expressed them. So I don't think we have anything to worry about
Bosnia and Herzegovina
Local time: 07:54
Member (2009)
English to Croatian
+ ...
Well, it never worked that way in history, in any industry. Birds didn't learn to fly from airplanes, but the other way round.
MT technology designers also compete between each trying too hard to prove it's their technology that works, etc etc. Also, the promotion is just too aggressive. Why don't you just relax and moni... See more
Well, it never worked that way in history, in any industry. Birds didn't learn to fly from airplanes, but the other way round.
MT technology designers also compete between each trying too hard to prove it's their technology that works, etc etc. Also, the promotion is just too aggressive. Why don't you just relax and monitor the natural language process, you may learn something from it.
[Edited at 2016-05-19 09:54 GMT] ▲ Collapse
Italy
Local time: 07:54
Member
English to Italian
That's the only explanation I can find to his "MT post-editing leads to lower quality translation – false. The translator is always free to ignore the MT just as he or she can disregard TM partial matches. Any effect on quality is probably due to priming, apathy, and/or other behavioral phenomena" in connection with the previous "MT is one scapegoat for ever decreasing... See more
That's the only explanation I can find to his "MT post-editing leads to lower quality translation – false. The translator is always free to ignore the MT just as he or she can disregard TM partial matches. Any effect on quality is probably due to priming, apathy, and/or other behavioral phenomena" in connection with the previous "MT is one scapegoat for ever decreasing per-word rates, especially among independent translators".
In fact, quite astonishingly, the author seems to ignore the small detail that many agencies (AFAIK) are exploiting MT in order to actively lower rates and that "ignoring the MT" means retranslating from scratch at those lower PE rates. If, on the contrary, a translator just does what they're being paid for, that is to say, just edit MT, then it will most probably result in awkward phrasing/syntax (which, IMO, is "lower quality translation"). ▲ Collapse
Local time: 00:54
German to English
+ ...
A section in the introduction uses a strategy that we see too often in these kinds of articles. "The choice to accept post-editing work is often cast in moral terms. " Anyone having an opinion contrary to the one desi... See more
A section in the introduction uses a strategy that we see too often in these kinds of articles. "The choice to accept post-editing work is often cast in moral terms. " Anyone having an opinion contrary to the one desired by the writer is cast as being irrational, emotional and similar - kind of a pre-ad hominem pre-attack I guess, putting words into mouths of people who have not weighed in. At the same time this strategy intimidates people from making statements, because they don't want to be seen as such silly irrational dinosaurs.
To set the record straight, my own position of post editing is neither emotional, nor moralistic (huh?!) nor irrational. Like any experienced trained professional I look at what is in front of me, and I also do so as a business person. I assess each project that comes my way individually. Editing and proofreading are always charged per hour, because poorly translated work will take longer to fix than well translated work of the same length, so this is fair for both the client and myself. So if a machine-translated text comes my way, I treat it like any translated text - I see how well it has been translated. If it is poorly translated, then fixing it will become more expensive at a per-hour fee than retranslating it would be, and I advise my client accordingly. Otoh, if the client wants to pay me 0.05/word when my translation fee is 0.15/word (French), and if it takes me twice as long to fix it, as it would to retranslate it, then this translates into a fraction of my usual per-hour earnings (divide total amount earned by hours needed to earn it). This is a combination of professional knowledge and business assessment. It is not emotional. It is not "moralistic". It is not irrational. So please do not paint us with that kind of a brush.
MT also gets misused. There is at least one designer/maker of MT who tries to educate his clients to that effect. The misuse often goes somewhat like this. You take a text of any kind and run it through MT, regardless of whether MT lends itself to that kind of text. Then you hire a translator to post-edit that text at a per word rate - say .03/word, so that a 1,000-word text will cost $30 instead of $100 or $150. It is a combination of greed and ignorance. A simple policy adopted by everyone of charging per hour instead of per word would fix this, and would make the practice discouraging for those clients wont to abuse that tool.
Btw, when the author cites the emotional reaction to people's mention of Google Translate when you say you are a translator, I have never had that said to me, ever, by anyone. Well maybe it's a fact of living in the capital of an officially bilingual country, where everybody's cousin, aunt, wife, or uncle seems to be a translator. ▲ Collapse
Denmark
Local time: 07:54
Member (2003)
Danish to English
+ ...
...
"MT is one scapegoat for ever decreasing per-word rates, especially among independent translators".
In fact, quite astonishingly, the author seems to ignore the small detail that many agencies (AFAIK) are exploiting MT in order to actively lower rates and that "ignoring the MT" means retranslating from scratch at those lower PE rates. If, on the contrary, a translator just does what they're being paid for, that is to say, just edit MT, then it will most probably result in awkward phrasing/syntax (which, IMO, is "lower quality translation").
-- my emphasis.
MT is used as an instrument for pressing word rates, and independent translators are often the small fish who find it difficult to put up any resistance.
Perhaps we should challenge the per-word rates and demand per-text rates, though this will be difficult. Not impossible, however.
Other professions would laugh you out of town if you suggested a per word rate for their work - contracts, reports, minutes of meetings, whatever. We need to start charging a rate for a job, primarily based on the time it takes and the qualifications required. We are NOT copy-typists, and we are not selling groceries by the kilo.
Professional translators will simply have to protest at every opportunity to agencies. Not necessarily to the PMs, whose hands are often tied, but by contacting the people who make the decisions. Make out a standard mail, so that it is quick and easy to post it off to directors and policy makers.
We will have to climb down from our ivory towers and emerge from our cosy back offices. It means making translators visible at trade fairs and school careers days, wherever we meet end clients and potential new translators, and discussing the issues.
The translator's job may have to be redefined, because machine translation is not going to go away. However, I am quite sure there is still plenty of work for human translators, and there are still clients who appreciate it and will pay for quality.
So-called machine translation is in fact heavily dependent on human translation. We translators need to be able to work with it and make sure we get our share of the advantages, both in money and in technical benefits.
Spain
Local time: 07:54
Spanish to English
+ ...
The author is not a translator himself and, what's more, he is the CEO of a company developing a new MT technology.
Yes, and, as we say in my home town, "it's ripping out of him"...
I do use MT, but judiciously. I'm actually more worried about online proofreading services, after receiving a spam email about them the other day. If used carefully, inept amateurs could use cunning a blend of limited language competence, MT and free proofreading services to turn out reasonably acceptable documents, thus further undercutting the market. After trying out one such service with a sentence from a text I was translating and having it returned unscathed, I received an email response:
"Thanks for trying out NAME OF PRODUCT! Did you know that you can use our service for free every day? You have one proofread per day, up to 140 chars, free of charge. And if you need to proofread more text, we have absolutely the lowest prices on the Internet..."
Spain
Member (2014)
English to Spanish
+ ...
When output from translation process is used by machines, we'll all be done here.
So long the output is used by humans, MT's got no chance to "outtranslate" us.
Spain
Member (2014)
English to Spanish
+ ...
"MT post-editing leads to lower quality translation – false. The translator is always free to ignore the MT just as he or she can disregard TM partial matches. Any effect on quality is probably due to priming, apathy, and/or other behavioral phenomena"
They forgot to mention:
"The translator is always free to ignore the rates offered by LSPs for post-editing. Instead, he or she can charge normal translation rate"
If we include... See more
"MT post-editing leads to lower quality translation – false. The translator is always free to ignore the MT just as he or she can disregard TM partial matches. Any effect on quality is probably due to priming, apathy, and/or other behavioral phenomena"
They forgot to mention:
"The translator is always free to ignore the rates offered by LSPs for post-editing. Instead, he or she can charge normal translation rate"
If we include this statement, I am more than willing to take on post-editing projects. ▲ Collapse
Local time: 13:54
English to Chinese
+ ...
The MT is sometimes useful and useless.
Poland
Local time: 07:54
English to Polish
+ ...
Two basic problems with that:
1. Clients and agencies tend to take a hostile or passive-aggressive attitude to departure from TMs. They sometimes try to appropriate the final word to themselves and are ready to make irrational decisions, i.e. decisions that favour and prioritize inaction.
2. Part of the context is financial — fuzzy discounts, not paying for repetitions etc. Cost-cutting trumps common sense (refer back to #1).
I don't want to speculate too wildly, but I sense an what I call the IT attitude to translation, in that comment. It regards humans as 'organics' or 'meatbags', which are typical anti-human slurs used by droids in science-fiction.
Humans are so bad, machines are better. Nope, just nope.
By way of conclusion I'll just comment on the final recommendation given by the author:
It could make sense for practical reasons, but nothing comes without a cost, including those things that come without a price.
Writing and editing are different skillsets and mindsets. PEMT-ing is just not the same as translation. From the perspective of progress of the human race, MT probably has a potentially significant role to play, but someone still has to supervise it, and the emotional costs of working as a full-time post-editor can be brain-breaking and soul-destroying.
And that could — to some extent — be mitigated with psychic hygiene and dignified human treatment. This means:
1. Non-antisocial working hours and working speeds, and workloads. Hence not the ultrafast conveyor belt large corporations would have in mind. Humans thrive when they work in human-optimized work settings; they can't beat machines in machine-optimized settings, just like machines can't compete in an inverted scenario.
2. No killing two birds with one stone and trying to lower translators' profile, aspirations and social class all while switching to MT. Keep it to MT integration and cut out the extraneous crap that serves your other agenda. Then we can talk about human translators embracing MT.
Else, MT is going to remain what it currently is, i.e. a symbol of all that is wrong with the 'industry'.
For the record, what's happening to the ATA also is such a symbol, with the ATA taking a marked pro-agency slant and including agency bosses among board members. From a translator's point of view this means the ATA has at least partially sold out and is no longer representing translators.
Poland
Local time: 07:54
English to Polish
+ ...
As translators, we have to stop competing with computers, because they win hands down at some things. We have to find the things they can't do, and so far there is still plenty of work for humans.
Yeah! You can't win too many battles if you don't get to choose the field. This is the hard side. The soft side is that we need to be more outspoken about the advantages of working with human translators in general and with freelance translators in particular, rather than being intimidated into allowing agencies to keep up their pretence of having more legitimate advantages and more important ones than is really the case. We need to increase visibility, and we need to make our voices heard. Faint voices are not heard.
Instead of taking natural translation processes and create technology from it, now they seem to create technologies first and then ask us to fit in them.
Spot on.
We will have to climb down from our ivory towers and emerge from our cosy back offices. It means making translators visible at trade fairs and school careers days, wherever we meet end clients and potential new translators, and discussing the issues.
As the owner of http://www.towerofivory.net, let me be allowed to offer a comment on that one as well.
Yes, we need more presence. We probably need to penetrate and infiltrate the market lest our ivory towers turn into bamboo cages and manholes.
On the other hand, we do need our ivory towers because we need to defend the legitimate right of our profession to respect commensurate with educational and skill requirements and the responsibility and scope of the tasks we do. We need to reclaim the credit arrogated to themselves as translation agencies that leaves only an empty shell to the translator, who is regarded as little more than a nameless data-entry employee regardless of how important texts and how well he or she has translated.
It's war, basically. You can have war with only minimum bloodshed, but the whole point of war is change of control of territory and especially strategic locations. You can't really wage war if you're too shy to claim territory, even though, of course, you shouldn't spread yourself too thin. But you don't really win wars through passive defence unless you're waging a war of attrition to preserve the status quo, e.g. mountain clans defending their inaccessible holdings from the encroachments of newly emerging central authority; exotic nations defending against costly invasions by Western powers. Here, we need to fight back to reclaim agreeable working conditions and rates and credit, so we can't just sit back and wait through frontal assaults. We need to go out and claim some ground, and hold it.
[Edited at 2016-05-24 14:15 GMT]
Japan
Local time: 14:54
Japanese to English
Actually, MT is a high-recall system for finding statistically similar segments from a large collection of corpora. MT is just a glorified TM: a TM's slightly smarter cousin.
MT still relies on translated text that it has "seen" (i.e., the corpora it's been trained on!)...but it goes a step further than a TM and uses statistical analysis to make educated "gues... See more
Actually, MT is a high-recall system for finding statistically similar segments from a large collection of corpora. MT is just a glorified TM: a TM's slightly smarter cousin.
MT still relies on translated text that it has "seen" (i.e., the corpora it's been trained on!)...but it goes a step further than a TM and uses statistical analysis to make educated "guesses" about unseen segments based on how similar they are to the segments it has seen.
One major danger: garbage in, garbage out. If the corpora an MT is trained on are full of errors or poor translations, then the results will be too.
And: the MT of today still doesn't know very much about context; it relies on the idea (more like, the "hope") that words/phrases that are similar to the ones it was trained on will have the same contextual relationships with each other and the surrounding text in the new/previously unseen segment.
Also: MT knows even less about anaphora (i.e. words, such as pronouns, that refer back to words or phrases used earlier in a text). See here for a short examination of some of the problems MT faces in dealing with them. In the meantime, here's a snippet.
Statistically, the 3 sentences above are very similar because they usemost of the same words, however the anaphor (it) used in each one refers to a completely different word or phrase, something that current MT is unable to account for.
My point is: MT still has a looong way to go before it can seriously compete, in terms of quality, with a real translator!
P.S. I tried "Lilt," one of the applications the author mentioned....The suggested translations constantly pop up as I type, and are both distracting and annoying. Instead of speeding things up, an app like that is only going slow me down! I think I'll pass... ▲ Collapse
To report site rules violations or get help, contact a site moderator:
You can also contact site staff by submitting a support request »
This discussion can also be accessed via the ProZ.com forum pages.