Should Professional Translators Do PEMT Work?
Thread poster: xxxsjmdcl
xxxsjmdcl
Local time: 12:17
English to Spanish
+ ...
Apr 4, 2013

You know, they say the work require light translation, requiring minimum effort from the linguist. This sounds very unprofessional to me and almost insulting. Will you ask a doctor to perform a light surgery which requires little effort of his/her part? What about asking an Engineer in charge of building a bridge to do only a light and quick calculation?

I think that professional translators can't be do light editing of machine translations because this is not professional.

What are your thoughts?


Direct link Reply with quote
 

José Henrique Lamensdorf  Identity Verified
Brazil
Local time: 13:17
English to Portuguese
+ ...
It's not light editing Apr 4, 2013

Let's assume that a professional translator would do a certain job for $100. Most likely, s/he'd review/proofread/edit/whatever it for $35.

While machine translation was under development, it was unable to provide anything barely acceptable. From what I've read and seen, MT has asymptotically reached the best level of performance it could. No breakthroughs to be expected in the foreseeable future.

Before MT reached its current development stage, the trick bottom-feeding translation outsourcers used to play was:
a) Translation (of the example above) by some amateur for $30;
b) Proofreading by the professional above for $35, as stated.

Bottom line is that they got a $100 quality translation for $65, and profited $45 more.

Of course, the pros they hired would have to be expendable. Nobody would accept 'proofreading' that $30 junk for $35 more than once.

But nowadays machine translation output, on this scale, is about as good (or as bad) as a $50 translation. Were that not enough, the $100 professional translator has raised his/her prices to $120/$40 respectively.

So the new trick bottom feeders are trying to play now is:
a) Translation by computer for $0;
b) Proofreading by that professional for $40.

Their desired bottom line is to get a $120 quality translation for $40, so they may pocket $80.

Will they make it?

Not with me. In the scale adopted for this exercise, my rate for PEMT is $120. They seem oblivious to the fact that if MT were any good, I could use it for free too, and then do PEMT instead of translating from scratch, to make more money.


Direct link Reply with quote
 

Karl Galgana Schimkowski  Identity Verified
Germany
Local time: 16:17
Member (2013)
English to German
+ ...
Google translate Apr 4, 2013

I suppose your question also refers to a recent job posting her on proZ. They basically give you a TM in which the source text appears as the target text and the program proposes a machine-translated version done by google. Your job would be to make only the bare minimum to make the sentences somewhat understandable.

There is no simple answer to the question whether it is ethical or not to do this because machine translation has been used before for texts that normally would not have been translated at all. In such cases, the machine translation serves the purpose of letting you get an insight into the content of the text. It is not (should not be!) intended to be published.

Nevertheless, there is a danger that clients will opt for such a translation to save money. If the client is told that it is just a rough translation (almost a word-for-word translation) like the kind used for a scriptwriter making a foreign language version of a screenplay for the dubbed version of films, and if the client is made aware of the stilistic inaccuracy, the translator/agency has fulfilled their responsibility. It is then up to the client only to peruse the translation for the intended use, i.e. getting a glimpse into the content. But as we all know, some irresponsible cheapskates don't understand the value of good translations and only see that they could save some money.


Direct link Reply with quote
 

Shai Navé  Identity Verified
Israel
Local time: 17:17
Member
English to Hebrew
+ ...
I completley agree with José Henrique Apr 4, 2013

The discussion about MT encompass various angles. The rate, although the most common one, is just one of them. Aspects like quality trade-off; the question of actual productivity gain in certain controlled environments (and here I refer to a trained "quality" TM engine and NOT to generic ones) compared to high quality work done by a professional with good TMs, Glossaries (that they have created and curated over the years), and expertise in the subject field; if widely adopted, the long term affect on the language and writing standards; the long term effect on the motivation of the translators doing the post editing work and its affect on the very essence of the marketplace, and much more.

Some of this is covered in a discussion on LinkedIn Ever accept post-editing work for machine translations?. I just hope that it is easily accessible to non-LinkedIn members or non group members.

[Edited at 2013-04-05 09:25 GMT]


Direct link Reply with quote
 

Łukasz Gos-Furmankiewicz  Identity Verified
Poland
Local time: 16:17
English to Polish
+ ...
Generally no but... Apr 5, 2013

Hard to say. I'm inclined to say no, forget it. And I never accept such jobs. I also thunder when parts of the source to be translated seem to be a machine translation from some other language. However, I remember an emergency situation proofreading translations worse than Google would have delivered. As I told the boss, it would have been better for us to skip those guys and just worked on Google's output between him and me. Thus PEMT seems to make more sense than proofreading some stuff you can get from live translators. But I don't agree with the use of that as a savings method where the competition is my own work paid for in full unless maybe for grammatically primitive jobs that come down to processing the vocab. Anyway, what matters is also the ratio of time invested to pay earned. For jobs where the quality is essentially the same but the translator earns more PEMTing than translating, I guess it's fair. But I don't think a professional translator should sit on a conveyor belt PEMTing. (Not that I necessarily know better than a translator who doesn't mind.)

Direct link Reply with quote
 

Heinrich Pesch  Identity Verified
Finland
Local time: 17:17
Member (2003)
Finnish to German
+ ...
So the amateurs loose Apr 5, 2013

Looks like MT is going to make amateur translators a thing of the past. Many of those amateurs turned later into pros, like me. This is a bad thing. Each of us has gone through a phase in our career when we were not so very good yet. If MT replaces the amateurs or start-ups, who will translate after 30 years? Hope MT will turn pro.

I use Google translate for some jobs. In technical standardised texts it works sometimes astonishingly well, but in free prose the output needs more editing than the translation from scratch. At least GT does not understand, that in German not all verbs in a sentence come at the end of the sentence.


Direct link Reply with quote
 

José Henrique Lamensdorf  Identity Verified
Brazil
Local time: 13:17
English to Portuguese
+ ...
Potential for a paradigm shift Apr 5, 2013

Łukasz Gos-Furmankiewicz wrote:

However, I remember an emergency situation proofreading translations worse than Google would have delivered. As I told the boss, it would have been better for us to skip those guys and just worked on Google's output between him and me. Thus PEMT seems to make more sense than proofreading some stuff you can get from live translators.
...
But I don't think a professional translator should sit on a conveyor belt PEMTing.


Heinrich Pesch wrote:

Looks like MT is going to make amateur translators a thing of the past. Many of those amateurs turned later into pros, like me. This is a bad thing. Each of us has gone through a phase in our career when we were not so very good yet. If MT replaces the amateurs or start-ups, who will translate after 30 years? Hope MT will turn pro.



My take on the situation is that a day may come when translation bottom-feeders will realize that the output they are getting from cheap amateurs, though different in the nature of its shortcomings, fails to surpass in overall quality that machine translation offers for free and in real time.

Sometimes machine translation is good enough for the intended purpose. The cost/benefit ratio should be allowed to decide that. I've put a side by side comparison example at http://www.lamensdorf.com.br/comparison.html for those who can read PT-BR.

What would happen to fledgling translators then? Where would they find a "vicarious learning ramp"? Well, technology has provided it already.

The new translator's training, on top of language skills, would include mastering one or more CAT tools. These are just pieces of software, so reading manuals and some exercising should cover it.

Then they would join a cloud-based translating team to handle those ominous 6-digit word count translation projects that recur in many industries, where repetitions and fuzzies abound. They'll be exposed to hundreds or thousands of professionally translated, half-translated, and untranslated segments, with the assignment of completing all of them to 'translated and proofread' status every single day. By sheer osmosis, in no time they'll become experienced translators in terms of volume (instead of time).

This could be the new fast track to professional translation.

The reverse path is tougher, and I tried it recently. My status: 39 years in professional translation, the last 9 of which using WordFast Classic.

I joined one such team around MemoQ, to which I was first introduced on that occasion. Pretty good software; it's in my plans to learn and switch to it in the near future. It's worth mentioning that the subject matter of those humongous translations was my main specialty.

The fact is that in spite of my specialization, which possibly brought some valuable terminology and conceptual contributions to the entire team, my lack of familiarity with the software, and not being an extremely keen WordFast user, led my productivity to be way below expectations. In spite of the PM being a personal friend as well, we decided that I'd better drop out of that team.

So in these CAT tools era, learning to use them before accruing actual translation experience seems to be a sensible paradigm. Learning translation from practice would be the next step.

This could have good consequences too. CAT tool developers, instead of muscling their way into translation agencies, could focus their marketing on the positive side. By offering free and high quality training to translators under development, they could eventually advertise their product as having more fully-trained translators in more language pairs worldwide than their competitors.


[Edited at 2013-04-05 11:09 GMT]


Direct link Reply with quote
 

Agnes Lenkey  Identity Verified
German to Spanish
+ ...
Yes, but at the full rate (=NO) Apr 5, 2013

Thank you all for your extremely interesting ideas and thoughts. My answer is yes, we may do PEMT, but at the usual rate – I know it sounds weird and is not a direct answer to the question, but for me it is the only possible way. Some direct clients (no agencies) tried to explain to me that they need proofreading of a text, because admittedly they paid a cheap translator and the outcome is not good enough/unusable for their purposes.
After having a detailed look at the original text and its translation, I always take my time (which I don’t have) to explain to the client that I am really sorry, but I have to charge my full rate – it is easier for me to translate from scratch than rewrite an unusable text, on the top of it being obliged to check the correspondence with the original text. What I do not explain to the client (because, at the end, my time has its limits, too) is that: a) either the translator was really amateur or b) maybe the translator cheated him/her and used MT instead of doing his work. At this point it makes no difference, the outcome is the same. So my answer is always: you can send me the original text together with the translation, I will be happy to help you – but I will have to charge you my full rate. In some cases I really had the feeling that the client was cheated by some unethical translator.

Thank you José for all your interesting thoughts. One made me really smile:

"Then they would join a cloud-based translating team to handle those ominous 6-digit word count translation projects that recur in many industries, where repetitions and fuzzies abound. They'll be exposed to hundreds or thousands of professionally translated, half-translated, and untranslated segments, with the assignment of completing all of them to 'translated and proofread' status every single day. By sheer osmosis, in no time they'll become experienced translators in terms of volume (instead of time)."

I agree with you, the reverse path is tougher, much tougher.. Sometimes, when using my "old" methods that involve paper, pencil and "classical memorizing" helped out by the simple search option of Word, I feel like a dinosaur. It has its good part, too, the satisfaction of being able to carry out a high quality translation only with the methods I was taught in a University where there were almost no computers at that time. Today I could not work without Internet anymore. Still, this whole new cybernetic world makes me feel very alone sometimes, having the feeling that I am always way back on this reverse path and never will catch up with all the others, the younger ones. It is a different path nowadays, like Heinrich says, but we always must adapt to reality, as best as we can.


Direct link Reply with quote
 

Thomas Rebotier  Identity Verified
Local time: 07:17
English to French
very dangerous for your skills Apr 6, 2013

I am a veteran of using MT as an inspiration. On short and plain sentences, you may have nothing to change. But the practice of "fixing" a long sentence with "light edits" will blunt your language sense. Unfortunately, this is what you are asked to do and are paid to do, the reason being that local changes will be picked up more easily by the 4-gram model of the MT.

Explanation why you are asked to be "light" : all current TM systems, including Systran now, are statistical or hybrid systems. The computations increase exponentially with how many words they correlate in the training set (the corpus of past translations used to tune the TM engine), so they usually have models up to 4 words at a time, maybe 5 for huge number crunchers like Google can afford. This means that if you change a long chunk of a sentence, you're not helping at all with the relations between the words that were wrong in the original translated sentence. You're providing another positive training point, but you're not giving corrective info to the system. Corrective info is precious, in fact there's a theorem on language learning that you cannot learn without negative feedback. So, fixing just a word or two is (1) cheaper but mostly (2) useful to improve the TM engine.

However, if you do that too much, you're also "retraining" your own language skills to the level of output that PEMT produces. As you all know, when a sentence is complicated, you ask yourself, "how would I explain this to my mom". With PEMT, you just leave you mom flustered. Don't train yourself to do this. It's a different job, better done by people who don't have fine writing skills to lose.

If you use MT, and I do, just remember that once you have decided to step in, you never erase enough. If your internal buzzer rings, read it for suggestions, then triple click the whole segment and hit delete. You'll be thankful on the long term.


[Edited at 2013-04-06 05:55 GMT]

[Edited at 2013-04-06 05:55 GMT]

[Edited at 2013-04-06 05:56 GMT]


Direct link Reply with quote
 

Karl Galgana Schimkowski  Identity Verified
Germany
Local time: 16:17
Member (2013)
English to German
+ ...
tolerating stylistic shortcomings Apr 6, 2013


However, if you do that too much, you're also "retraining" your own language skills to the level of output that PEMT produces. As you all know, when a sentence is complicated, you ask yourself, "how would I explain this to my mom". With PEMT, you just leave you mom flustered. Don't train yourself to do this. It's a different job, better done by people who don't have fine writing skills to lose.


This is also what I believe. I saw guidelines for PEMT proofreading, stating that you should "ignore stylistic issues" and that things like coherence, cohesion, standard word order etc. are "less important". Any language professional will risk having a bad conscience sending in a text with such deficiencies as finished work.


Direct link Reply with quote
 


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


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

Should Professional Translators Do PEMT Work?

Advanced search







PerfectIt consistency checker
Faster Checking, Greater Accuracy

PerfectIt helps deliver error-free documents. It improves consistency, ensures quality and helps to enforce style guides. It’s a powerful tool for pro users, and comes with the assurance of a 30-day money back guarantee.

More info »
Wordfast Pro
Translation Memory Software for Any Platform

Exclusive discount for ProZ.com users! Save over 13% when purchasing Wordfast Pro through ProZ.com. Wordfast is the world's #1 provider of platform-independent Translation Memory software. Consistently ranked the most user-friendly and highest value

More info »



Forums
  • All of ProZ.com
  • Term search
  • Jobs
  • Forums
  • Multiple search