Pages in topic:   < [1 2 3] >
Machine translation and price reductions
Thread poster: Thomas Loob
Kay-Viktor Stegemann
Kay-Viktor Stegemann
Germany
Local time: 15:48
English to German
In memoriam
About fairness Sep 21, 2020

Thomas, a discussion about "fair" rates is rather moot. Fairness is a category that applies to wages: if you are employed, you need fair wages to make a living, and this kind of fairness is mostly enforced by local governments, for example by setting a minimum wage, or by unions, by negotiating some kind of collective agreement.

This kind of protection does not really exist for freelancers, particularly not for those who offer their services on the world market. It is impossible to
... See more
Thomas, a discussion about "fair" rates is rather moot. Fairness is a category that applies to wages: if you are employed, you need fair wages to make a living, and this kind of fairness is mostly enforced by local governments, for example by setting a minimum wage, or by unions, by negotiating some kind of collective agreement.

This kind of protection does not really exist for freelancers, particularly not for those who offer their services on the world market. It is impossible to enforce any kind of collective agreement or law if the next competitor in a low wage country can simply ignore both, since it does not apply to them. Like it or not, you cannot eliminate the market mechanism of supply and demand. Consequently, if you want to thrive as a freelancer, you need to be in a position where demand is high enough and supply is limited. If you succeed in moving into such a position, no one can force you to accept any sort of discount for whatever pretext. If you are not in such a position, the best idea would be to look for ways to improve your position. That is the only real chance you have. Even the most benevolent client or agency will not offer you better rates just for the sake of fairness, particularly not if the next translator is prepared to do the job for less. However, they will, and they must, if the supply of translators who can deliver what they need is limited.

I suspect that rates for MTPE and the like will continue to erode, simply because MTPE is a declaration of "we don't need that last 5% of quality for our purpose, let the reader deal with it". That allows competitors from low wage countries to offer MTPE too, for less. Look out for clients and jobs where this last 5% of quality is essential, and you find a market where competition is moderate and where the professional translator will be allowed to choose their own tools.
Collapse


Erik Freitag
Mina Chen
Kuochoe Nikoi-Kotei
Maria Teresa Borges de Almeida
Arkadiusz Jasiński
Christopher Schröder
Germano Matias
 
Paul Adie (X)
Paul Adie (X)  Identity Verified
Germany
Spanish to English
+ ...
General comment Sep 23, 2020

As time goes on, I see more post-editing work, which I don't do myself, but I wonder how long the translation industry can last as it has done. I have tested several MT engines, and the fact is - they do get the general meaning across. Of course, for published work, medical and legal translation, philosophy and a long list of etceteras, a human translation will always be necessary, but for a large amount of commercial translations, I see MT being the first port of call for many.

I w
... See more
As time goes on, I see more post-editing work, which I don't do myself, but I wonder how long the translation industry can last as it has done. I have tested several MT engines, and the fact is - they do get the general meaning across. Of course, for published work, medical and legal translation, philosophy and a long list of etceteras, a human translation will always be necessary, but for a large amount of commercial translations, I see MT being the first port of call for many.

I was trained and started out just before MT really started taking over and am grateful for this, but apart from very specialised areas, I see MT becoming more and more popular in the future (also for interpreting). Just saying "no, this is not how it is done" doesn't help anyone. The output of MT has helped me "understand" texts written in Chinese, Japanese, Greek and even to "talk" in Chinese!

However you feel about it, it's here to stay. I do hope that it doesn't take over the work I do, but I wouldn't be surprised if it started to gnaw at some point.

Wishing you all well,

Paul
Collapse


Christopher Schröder
Jorge Payan
Ingvild Karlsen
 
Kay-Viktor Stegemann
Kay-Viktor Stegemann
Germany
Local time: 15:48
English to German
In memoriam
Commercial translations Sep 23, 2020

Paul Adie wrote:
As time goes on, I see more post-editing work, which I don't do myself, but I wonder how long the translation industry can last as it has done. I have tested several MT engines, and the fact is - they do get the general meaning across. Of course, for published work, medical and legal translation, philosophy and a long list of etceteras, a human translation will always be necessary, but for a large amount of commercial translations, I see MT being the first port of call for many.
Paul


There are a lot of commercial translations where MT does not really work, for example:
- technical translations. A user guide for heavy equipment needs to be translated by a human if you do not want to risk live and limb, and even below these threats and for simpler appliances, any number of things will go wrong if terminology is not applied correctly and in a context-aware way. Take the term "socket" for example: this can mean a kind of screw, a kind of wrench to tighten that screw, a wall outlet, a connector socket, or any number of other fitting-, sleeve- or plug-like thingies. Applying the right term here is essential and at the same time practically impossible for MT, and there are endless similar examples. You simply need someone (read: a person) who understands what they are talking about, and can look at a picture or ask for clarification if in doubt. In many cases, the use of MT will cause more work than it saves, if any.
- marketing translations. Even if your statement that MT "gets the general meaning across" were true (which depends on how much slack you cut the "general" here), getting the meaning across is often only a small part of the work. Many commercial texts are about selling something. And this does not only apply to hard marketing copy, it applies to any published text by or about a company or product, be it their website, a press release, a software description, a training video or whatnot. All of this is marketing, or contributes to it. All of this can make or break the reputation of a company, if the text is boring, long-winded, stiff, repetitive, or lacks credibility because it is not adapted to the target audience, even if it "gets the general meaning across". Again, getting this right mostly means that you have to discard the MT result and do a rewrite in an attractive, compelling and convincing language.

Of course, there is a market for mediocre commercial translations. As there is a market for any kind of mediocre products. There will always be companies who offer cheap products or services and try to increase their international market share by quickly adding translated product content, without much regard for quality. That's what "light PE" is for, if they don't do it by MT alone and damn the consequences. That market is mostly lost for professional translators, except for the cases where the company suddenly finds that it loses money with its mediocre translations: by losing sales, reputation, or even lawsuits.

Instead of trying to supply this low end market, professional translators should focus on the market where clients are aware of the consequences of poor output. Or where we could make them aware of it.


Aisha Maniar
Sandra & Kenneth Grossman
Daryo
 
Thomas Loob
Thomas Loob  Identity Verified
Estonia
Local time: 15:48
Member
English to Swedish
+ ...
TOPIC STARTER
Discussion off the mark Sep 23, 2020

My initial aim was to discuss price reductions due to MT.
It is about editing MT into good text with highest quality, like it was 100 % human translated.
I have come across human translations that are as bad as MT.
MT is here to stay and some of the engines are very good now. Of course depending on the subject, complexity and languages.
We all know that MT is bad and subject to fear and and hate etc.
So why not create a trade union of all translators to resist it?... See more
My initial aim was to discuss price reductions due to MT.
It is about editing MT into good text with highest quality, like it was 100 % human translated.
I have come across human translations that are as bad as MT.
MT is here to stay and some of the engines are very good now. Of course depending on the subject, complexity and languages.
We all know that MT is bad and subject to fear and and hate etc.
So why not create a trade union of all translators to resist it?
Don't cry, organise, someone said...so please stop complaining and come up with constructive ideas.

[Redigerad 2020-09-23 10:00 GMT]
Collapse


Hannah Doyle
 
Philippe Locquet
Philippe Locquet  Identity Verified
Portugal
Local time: 14:48
English to French
+ ...
Backtranslating Sep 23, 2020

Thomas Loob wrote:

Your assumption that the material from MT engines are taken from the net is probably not correct as much content is already translated by bad MT engines, at least in my language.


Hi Thomas, this is an interesting topic you bring here.
It's nice to read comments from people that actually have experience dealing with real-word MT scenarios (such as Elena Feriani, Samuel Murray etc.), it's refreshing. As you've seen, MT as a topic is a bumpy ride.

Your suggestion touches on Ethics more than on practicality. Tracing where the training data comes from is an extremely difficult proposition. I'm guessing you're familiar with bilingual data such as TMs, the reality is that MTs are not trained exclusively on this.
MTs training material is called corpus. It can be:
_Monolingual
_Bilingual
_Glossary style

It can be obtained by:
_adding free bilingual material found for example on the opus website
_Buying bilingual material from human translation (type of purchase rare and expensive)
_Merger: a company buys another and accesses human translations this way and thus trains MT engines
_Alignment (seldom)
_Web scraping
_Backtranslating (to oversimplify: MT produced corpus)

Backtranslation seems to be getting popularity lately for that purpose. I suggest you search what I mentioned above so you can get a clearer picture of what's happening.

Note that MT programmers also create "tool" MTs which purpose is to test and correct the output of the engine being trained.
So the process of working on an MT engine to train it before it is ready for production is far more convoluted that it looks.

Hence when it comes to pricing, its is probably better for the translator to adapt the rate to the quality. The easiest way and probably fairer way that has been proposed on forums here is to charge by the hour. If the engine is crude, it will take longer to edit, which will be reflected in the price the translator charges.

My two cents


 
Christopher Schröder
Christopher Schröder
United Kingdom
Member (2011)
Swedish to English
+ ...
I think this is moot Sep 23, 2020

Thomas Loob wrote:

As many of you probably have experienced, machine translation (MT) is now so acceptable that translators are forced to accept discounts.
Have anyone argued that since the translator(s) have actually contributed to the mass of examples for the AI algorithms to chew on, that the amount of doscount should be shared? Example: if the discount is 25 %, it should be negotiated to, let's say, 12,5 %.


A translator (or editor in this case) is paid for their labour. The unit of labour is time. A price per word is just an approximation of the price per hour. So it makes no difference who discounts what. You just charge the equivalent of what you feel you are worth per hour. The quality of the MT determines how long it takes and so the price. Any “discount” will reflect time saved and have zero impact on you.

(Plus: nobody can be forced to do anything, and I have not even been asked.)

[Edited at 2020-09-23 11:22 GMT]


MollyRose
Jorge Payan
Robert Carter
 
Hannah Doyle
Hannah Doyle  Identity Verified
France
Local time: 15:48
French to English
+ ...
MT and price reductions are the future Sep 24, 2020

Yes, MT is on the rise. And yes, depending on the tools used, the machines are (unfortunately, disappointingly) doing a pretty good job - even in fields that would have been 'impossible' to MT a couple of years ago. The other day I saw a fair translation of a marketing text. Sure, it needed editing, but it was surprisingly smooth. Better than a shoddy human translation. I've been seeing more agencies move this way, and what has particularly saddened me, is seeing a boutique agency that used to g... See more
Yes, MT is on the rise. And yes, depending on the tools used, the machines are (unfortunately, disappointingly) doing a pretty good job - even in fields that would have been 'impossible' to MT a couple of years ago. The other day I saw a fair translation of a marketing text. Sure, it needed editing, but it was surprisingly smooth. Better than a shoddy human translation. I've been seeing more agencies move this way, and what has particularly saddened me, is seeing a boutique agency that used to give me long, sink-your-teeth-into-it texts for years, having been bought up. Now most of what they're asking for is PEMT, and I rarely work with them anymore.

It seems inevitable that the future lies in MT and massive price reductions. Perhaps not immediately, but in the next three to four years, sure. The post-covid economic crisis will undoubtedly precipitate things. Last Friday, my favourite client (a one-man agency) asked if there was any way I could translate 60k words for Wednesday - I said I could take part of it. He went away to find other translators to pair me with, and then came back and said the client had pulled the project - they'd decided instead to go with a UK agency that would MT it and post-edit in just 24 hours.

With everyone, everywhere, demanding ever faster results, how could translation be any different?

Eight years ago, I did my Masters in translation, where they talked about MT as some kind of threatening far-future thing to be vaguely aware of. Something for the next generation to deal with. It only took eight years for MT to become very visible. I think we have lots of change coming our way, faster than we expect. I'm not trying to be bleak or cynical, just realistic. I think the near future for translators will be learning to accept PEMT as a new facet of your work, or moving sideways and diversifying your business in a slightly different direction. Even clients that 'don't work like that', or projects that are 'unmachineable'... I just suspect massive advances in what the technology is capable of, and good non-MT-using clients will be forced to adapt. If an agency that pays its translators decently and works at a fast albeit human pace is up against agencies using high-performing MT and good editors, working at half the speed and half the price, that first agency will no longer be competitive.

For the time being, it's acceptable to stick to the old way, because there's still doubt surrounding MT, and some end clients don't trust the tech. But as I say, as the tech sharpens, the doubt will dissipate.

I find it really sad - but inevitable. Obviously this is just my opinion, and none of us can really know.
Collapse


Alvaro Palmieri
Philippe Locquet
Ruchira Raychaudhuri
 
Samuel Murray
Samuel Murray  Identity Verified
Netherlands
Local time: 15:48
Member (2006)
English to Afrikaans
+ ...
An alternative to the discount model Sep 24, 2020

Thomas Loob wrote:
Have anyone argued that since the translator(s) have actually contributed to the mass of examples for the AI algorithms to chew on, that the amount of discount should be shared?


I agree that the discounts for MT are very, very unfair. Here's why:

When I use Google Translate on my own, I pay about 0.2c per sentence. If the client supplies me with a machine translation, it saves me from having to pay 0.2c for every segment, so a fair discount should be 0.2c per segment. [1] Instead, clients insist on double-digit percentage discounts, which at a translation rate of 10c per word works out to about 50c per segment.

Now, of course, Google is a very large company and they can afford to sell their machine translations very cheaply. If the client created his own machine translation system and spent money on training and developing it, the cost per character may be quite a bit higher, but I doubt that it would cost as much as 50c per segment to defray.

On the other hand, it may be argued that the client's own machine translation system is likely to be less good (therefore less valuable), or even: since it is the client's own decision not to use Google Translate, any cost exceeding the regular price of Google Translate should surely be met by the client.

Now, with regard to your original suggestion:

If we were to apply a model of "buying machine translations from the client" and thus giving the client a discount per character, word or sentence based on the actual buying cost of the machine translation, we could also apply such a model for the licensing of translated sentences back to the client. (The regular translation rate includes a license for use for the client's regular purposes only -- I'm talking about an additional licensing fee charged for using our translations for the purpose of training a machine translation system, assuming for argument's sake that we can enforce such a licence.)

So, theoretically, it could work. But here's the rub: for just how much should you be selling your sentences back to the client? How much worth more is your sentence than the seeding sentences the client had originally purchased to train his machine translation system? How much value does your one sentence truly add to the quality of future machine translations produced by the client's system? I suspect the answer is: preciously little -- probably a fraction of the cost of buying a machine translation sentence from the client in the first place.

This means that even if the idea is good, the actual money to be earned on the additional licensing fees may be less than one cent per segment.

--
[1] (Some clients might argue that taking care of sending the source text to the machine translation system and automatically inserting it into the target field for you is worth something, but in reality it doesn't take more than 5 seconds per segment to do that manually, so the saving is negligible.)

[Edited at 2020-09-24 11:22 GMT]


 
Peter Motte
Peter Motte  Identity Verified
Belgium
Local time: 15:48
Member (2009)
English to Dutch
+ ...
I's spreading false opnions Sep 24, 2020

Thomas Loob wrote:

1. If you are not more or less forced, why discuss it when it doesn't apply to you?


The problem is that it does spread the idea that you should give a discount, and that people might be inclined to give, and thereby lowering the prices.
What some people forget, is that in the end it will lower the prices for everybody, ALSO FOR THE AGENCIES WHO SELL TRANSLATIONS. So, finally, everybody will make less money.


 
Peter Motte
Peter Motte  Identity Verified
Belgium
Local time: 15:48
Member (2009)
English to Dutch
+ ...
Fair is overrated Sep 24, 2020

The idea of a "fair" price is overrated.
All those agencies wanting you to give a discount, are not interested in being paid a fair price themselves, but in making as much money as possible.
So why would a translator settle for a "fair price" instead for "a much as possible"?
"Fair" turns easily into an abused moral term.

[Edited at 2020-09-24 12:13 GMT]


Jorge Payan
Daryo
 
Thomas Loob
Thomas Loob  Identity Verified
Estonia
Local time: 15:48
Member
English to Swedish
+ ...
TOPIC STARTER
? Sep 24, 2020

What do you think my post was aiming at in the first place?
Easy for companies to force discounts like 30 % for translating with MT when we are not joined together and lack solidarity.


 
Christopher Schröder
Christopher Schröder
United Kingdom
Member (2011)
Swedish to English
+ ...
Still confused Sep 24, 2020

Thomas Loob wrote:
Easy for companies to force discounts like 30 % for translating with MT when we are not joined together and lack solidarity.


Where is the problem in an agency paying you 30% less for a job that takes 30% less time?


 
Dan Lucas
Dan Lucas  Identity Verified
United Kingdom
Local time: 14:48
Member (2014)
Japanese to English
Again with the union idea? Sep 24, 2020

Thomas Loob wrote:
Easy for companies to force discounts like 30 % for translating with MT when we are not joined together and lack solidarity.

They can't force you. Just say no to MT and no to the company asking you for a discount. If they need your services they'll offer more. If they don't need them they won't. Easy. If you can't find clients prepared to pay for your services, you'll have to find another line of work.

And we've had this "let's unionise" thing more than once on this forum (do a search). The argument is not new and neither are the counter-arguments. This is a global, deregulated industry in which there is no requirement for a physical presence in the country where the work arises.

If you try regulating or unionising in one jurisdiction, many clients will simply stop dealing with freelancers in that state or country.

You'd think over the past 12 months freelancers would have learned something from AB5 in California about how easy it is for well-intended regulation to have material adverse effects on those it purports to help. There's a thread on that, funnily enough.

Dan


Kay-Viktor Stegemann
Jorge Payan
 
Susan van den Ende
Susan van den Ende  Identity Verified
Germany
Local time: 15:48
English to Dutch
+ ...
Not true per se Sep 24, 2020

Peter Motte wrote:

The problem is that it does spread the idea that you should give a discount, and that people might be inclined to give, and thereby lowering the prices.
What some people forget, is that in the end it will lower the prices for everybody, ALSO FOR THE AGENCIES WHO SELL TRANSLATIONS. So, finally, everybody will make less money.


This isn't true though, is it?

I offer a discount because I work with a CAT tool. But if that CAT tool allows me to translate 1000 words per hour rather than 400, and my discount is a whopping 50%, I still end up invoicing 25% more per hour than when don't use a CAT.

Same for MT: as long as the discount you offer is smaller than the benefit you get from a tool, you are increasing your earnings. So look at the bottom line:

- how much can you invoice for an hour of work?
- are you still enjoying that work enough to keep doing it?


Christopher Schröder
Hannah Doyle
Robert Carter
 
Daryo
Daryo
United Kingdom
Local time: 14:48
Serbian to English
+ ...
That IS part of this thread ... Sep 29, 2020

Thomas Loob wrote:

If you cannot discuss the question I raised, why not start another thread?


First you affirm (baselessly, as far as I'm concerned)

"machine translation (MT) is now so acceptable that translators are forced to accept discounts."

"Acceptable"? That might well be the opinion of some penny pinching "providers of linguistic services" who have only a passing / vague / at arm's length acquaintance with linguistics, and some clients who speak only their own language and take the hype (not to use a more accurate term ...) about the "marvels of Machine Translation" as gospel, but I don't know of any professional linguist who would say that.

Then when you don't like what you hear you "want everyone else to stick to the question" ...

Well FYI if a question is based of false premises, THAT needs to dealt with first - and ***is*** part of "answering the question".

MT is supposed to be acceptable for what exactly?

To be considered as a usable / reliable "translation"?

Possibly one day in some distant future, but definitely not "NOW".

Did you mean that MT is acceptable for "editing" - yeah - sure thing - definitely is!

You could call "editing" absolutely anything - from having to change few punctuation marks up to deciding that it's quicker to simply ignore the MT output and translate the ST yourself from scratch.

If your idea of "editing MT" includes (the most likely case, BTW) doing yourself the translation from scratch, then yes - any MT would be "acceptable" .. as much "now" as it was around 60-70 years ago.


 
Pages in topic:   < [1 2 3] >


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


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

Machine translation and price reductions







Trados Business Manager Lite
Create customer quotes and invoices from within Trados Studio

Trados Business Manager Lite helps to simplify and speed up some of the daily tasks, such as invoicing and reporting, associated with running your freelance translation business.

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 »