Pages in topic:   < [1 2]
Adaptive machine translation is coming? Or it's already here?
Thread poster: Spence Green
Spence Green
Spence Green  Identity Verified
United States
Local time: 12:12
Member (2014)
English to Arabic
+ ...
TOPIC STARTER
Of course you're right! Aug 24, 2016

Michael Joseph Wdowiak Beijer wrote:

I suspect there might be a few more variables we can control.

Michael


Yes, you're right, of course! We find that per-word rate discussions generate considerable enthusiasm. Less so, discussions about productivity. But one must contend with globalization when addressing the former, while there are more local solutions for the latter.


 
Bernhard Sulzer
Bernhard Sulzer  Identity Verified
United States
Local time: 15:12
English to German
+ ...
What translators can and should control Aug 24, 2016

Spence Green wrote:

Mirko Mainardi wrote:

On the other hand, we(*) believe MT, PEMT, adaptive MT and the like will "catalyze" even worse work conditions and rates for translators at large and contribute to further devaluing their profession and creativity.



This is a reductive analysis of downward pressure on per-word pricing in the industry. At least as powerful as technology are the low barrier to entry to the industry, outsourcing by agencies to low-wage countries, and the rise of commoditized translation platforms (Gengo, Unbabel, etc.) that emphasize convenience.


Low barrier to professional translating? Not really, you've got to have a few pretty good skills.
Low barrier to doing funny stuff? Yes, no doubt about it.

I am all for free market forces but in a highly sophisticated market as that of professional (not fly-by-night) translations, market participants are first and foremost interested in quality services from excellent service providers, i.e. agencies and translators or translators directly. Those simply don't come cheap.

That is the only market I am interested in because I am not willing to sell myself short, and I don't have too.
If I were literally forced to accept pittances for what I do, that would be proof that you can simply forget about being a professional translator and I would simply move on to greener pastures. Not all is green where I sit, but I am not willing to compromise my work for cheap pay; and I am increasingly finding clients and agencies that believe in the same principles as I do.

So, I can only speak for myself but I am not too worried about the Gengos and Unbabels, low-wage country competition and the low barrier to entry to the industry.
I am more worried about getting good technical tools that mean what they say and that for a fair price.

Spence Green wrote:
Absent collective action, freelance translators have no control over these three market forces.

The single variable that translators can control is hourly productivity and thus the *effective* hourly rate. And that is where automation---be it TM, MT, or any other linguistic resource---is obviously useful.


I don't need to have control over the three forces (as "market forces") you mentioned because they simply don't pertain to me in a way that would force me to give unjustified discounts or have to change my hourly rates/flat prices etc. downward. These factors haven't killed my business. Quite the opposite. I do control all aspects of my own productivity though.

As far as hourly productivity is concerned, it's not true that CAT tools or MT engines always speed up all your work and it's not like there are often many problems with CAT tools that I can't believe still happen although they've been happening for many years (from being unable to save target files with coded error messages that no layman or woman and often no expert can explain and you find out later that there are actually easy solutions; spell-checkers that correct like "they" speak a different language; no matching of segments although it's even obvious that a matching segment should apply, even at a 70% matching limit; the funny way TMs are saved automatically at various places but you better always make sure you do open them yourself and the right version (the one saved with the text); the fact that I can't just type quotation marks in the target language format and have to instead click on a symbol and then edit in or edit out the actual sentence; formats that just won't be tackled without a myriad of tags so you have to create simpler text versions which you then have to copy and paste back into the original format with tables and pics; to mention just a few from my own experience).

And if you are fast and can deliver within a very impressive time, you are the one controlling the price and that doesn't mean it's unfair NOT to discount - the opposite seems much more logical.

So what is it I control: the way I go about my business in a professional way, and that will yield fair rates, now and in the future.


Don't get me wrong, I like CAT tools. But I do not like hype or blind belief in technology.
Great productivity is and should never be an excuse or the reason for low prices; not in our business.
Technology helps but you need to know how to use it effectively. And that does not spell discount, especially not when some CAT tool arrives at a fuzzy word count. Nothing fuzzy about it.


[Edited at 2016-08-24 22:27 GMT]

[Edited at 2016-08-24 22:35 GMT]


 
Spence Green
Spence Green  Identity Verified
United States
Local time: 12:12
Member (2014)
English to Arabic
+ ...
TOPIC STARTER
Got it Aug 24, 2016

DLyons wrote:

If you actually had any respect for her, that pseudo-quote would not have been on your website!



Oh, I see. The text isn't in quotation marks, so there isn't any attribution, at least according to journalistic conventions? And it should be pretty obvious that someone who died a century ago wasn't a Lilt user....

As far as I can tell, these ads are no different in form than e.g., the "Got Milk" ads that feature celebrity faces, or Nike ads that say "Just do it" next to athletes' images, or Apple ads that say "Think Different" next to Gandhi, etc. I think that it's widely understood that Gandhi didn't have a Macbook.


 
Spence Green
Spence Green  Identity Verified
United States
Local time: 12:12
Member (2014)
English to Arabic
+ ...
TOPIC STARTER
Agreed! Aug 24, 2016



Low barrier to professional translating? Not really, you've got to have a few pretty good skills.
Low barrier to doing funny stuff? Yes, no doubt about it.

I am all for free market forces but in a highly sophisticated market as that of professional (not fly-by-night) translations, market participants are first and foremost interested in quality services from excellent service providers, i.e. agencies and translators or translators directly. Those simply don't come cheap.

That is the only market I am interested in because I am not willing to sell myself short, and I don't have too.
If I were literally forced to accept pittances for what I do, that would be proof that you can simply forget about being a professional translator and I would simply move on to greener pastures. Not all is green where I sit, but I am not willing to compromise my work for cheap pay; and I am increasingly finding clients and agencies that believe in the same principles as I do.



Agreed on all points.


As far as hourly productivity is concerned, it's not true that CAT tools or MT engines always speed up all your work and it's not like there are often many problems with CAT tools that I can't believe still happen although they've been happening for many years (from being unable to save target files with coded error messages that no layman or woman and often no expert can explain and you find out later that there are actually easy solutions;



I agree. See my previous comment on this thread about technology vs. rate discounting.


And if you are fast and can deliver within a very impressive time, you are the one controlling the price and that doesn't mean it's unfair NOT to discount - the opposite seems much more logical.


If I've resolved the scope of these negations correctly, then I think that you're saying if you can translate faster, then you can/should charge more. Sure: command the best price that you can! But another rational strategy is to gain market share by reducing the price.


Don't get me wrong, I like CAT tools. But I do not like hype or blind belief in technology.
Great productivity is and should never be an excuse or the reason for low prices; not in our business.
Technology helps but you need to know how to use it effectively. And that does not spell discount, especially not when some CAT tool arrives at a fuzzy word count. Nothing fuzzy about it.


Your price is a component of your competitive strategy, of course. Technology is another. I guess that you have to work backwards from your business objectives.


 
DLyons
DLyons  Identity Verified
Ireland
Local time: 19:12
Spanish to English
+ ...
Just one casual visitor's reaction Aug 25, 2016

Spence Green wrote:

DLyons wrote:

If you actually had any respect for her, that pseudo-quote would not have been on your website!



Oh, I see. The text isn't in quotation marks, so there isn't any attribution, at least according to journalistic conventions? And it should be pretty obvious that someone who died a century ago wasn't a Lilt user....

As far as I can tell, these ads are no different in form than e.g., the "Got Milk" ads that feature celebrity faces, or Nike ads that say "Just do it" next to athletes' images, or Apple ads that say "Think Different" next to Gandhi, etc. I think that it's widely understood that Gandhi didn't have a Macbook.



Yes, it's pretty obvious - and probably most visitors won't even notice. But my immediate reaction was very strongly negative, to the point that I dismissed out of hand what might be a useful tool.


 
Soonthon LUPKITARO(Ph.D.)
Soonthon LUPKITARO(Ph.D.)  Identity Verified
Thailand
Local time: 02:12
English to Thai
+ ...
Benefits of adaptive MT? Aug 25, 2016

Mirko Mainardi wrote:

No, really, Spence, translation is not assembly line work, not even in more "technical" fields (where there arguably is more standardization in terminology/syntax). And even if it were, in what universe this new and wonderful technology would only benefit translators, without agencies taking advantage of it to push rates further down (as they did/do with fuzzy matches and PEMT, for instance)?


I partially agree with Mirko. However, I concentrate on translation and interpretation of technical texts, I find that latest software and MT help me to work more efficiently. I beg not to ignore this technology when we are fighting back against the downward pricing of translation jobs.
For instance, MT makes us work quicker and need less human intervention to finalize large projects in a few day. In particular, PEMT is now standardized by ISO and other international standard institutions for global use.

Soonthon L.


 
Jeff Allen
Jeff Allen  Identity Verified
France
Local time: 20:12
Multiplelanguages
+ ...
The MT postediting ISO standard Aug 25, 2016

Soonthon LUPKITARO(Ph.D.) wrote:
In particular, PEMT is now standardized by ISO and other international standard institutions for global use.
Soonthon L.


Hi Soonthon,
The draft (ISO/DIS 18587.2) of the standard "Translation services — Post-editing of machine translation output — Requirements" is very generic. It reuses the entire concept of human resources verification of translators from (ISO 17100:2015) "Translation services -- Requirements for translation services" (originally the European EN 15038:2006 standard, and just adapts it to the role of a post-editor.
That part is a very simple checklist which heavily relies on the same requirements as for translators in the translation services standard. It really only adds another generic statement about amount of experience in post-editing. There is no such thing as approved or recommended post-editing training and certification programs because the type and amount of post-editing depends on the MT system which is used, the type and amount of upfront training of it, the language pair, and many other factors.
If I sat down with all of the translation post-editing teams worked with over the past 20 years, explained to them how an audit is conducted, coach them for 1/2 day and run some mini-check audit sessions with them, all of them could be certified next week. This covers a wide range of MT system approaches, different MT brands, completely different workflow processes, a variety of post-editing feedback channels, etc.

Unlike translation services, and even the use of other mature translation tools in translation workflows, this post-editing standard will face a much different challenge in the auditing cycle. This challenge is the notion of too many different ways to do the same thing.
After spending the past 10 years in a secondary role as an auditor of ISO9001 processes in software companies, I spend much time coaching different software development teams, especially acquired companies, to always refer to expected and stated workflows and processes of the overall company. Most of the weaknesses and non-conformities raised have been due to teams thinking and functioning in their silo without considering how their processes differ from those of teams in another part of the building.
An auditor will usually request to audit different teams and projects and wants to see if there is coherence across them within the company. And some companies even state in their documented processes the specific tools to be used. I often advise companies to avoid this because it then locks all teams into having to show and tell how they use that tool.
Let's take for example the idea of large translation supplier X which says that they use Trados Studio for all translation workflows. Yet, during an audit, the auditor discovers that the files received from freelance translation Y or Z are coming from MemoQ or Wordfast or something else. It is for this reason that one should never state a given too, but show how the process is done in using the various tools, and explain why. Always have things documented beforehand to show the various workflow differences. This is often now covered given that translation production teams often have licenses for the different TM/CAT tools to handle import and export issues.

As for MT, it is not so simple. To be able to use a statistics-based MT system, it requires having much content upfront to train the system. This is not always possible for each customer, and especially for each language.
On the other hand, rule-based systems are quite good for small projects of a 5000 - 500,000 words to quickly analyze the translation project, extract out all terminology, quickly create a customer dictionary with variants, and apply it immediately in combination with the TMs. It is possible to do all of this, including on-the-fly changes in real-time post-editing mode.

Some larger companies opt for using different MT system types and brands for different language pairs and also depending on the size of the projects.

In the case of such a variety of MT system and workflows available, it is very important to have each of them documented (even just visually with a minimum amount of explanations), so that a process auditor doesn't get confused and start noting weaknesses all of the place and label it with a non-conformity.

I wrote an article "Translation standards - which one?" for a translation publication at the end of last year. Maybe I should write a follow-up article on how to prepare for audits under the upcoming post-editing standard in order to avoid receive negative points by auditors.

Jeff


 
John Fossey
John Fossey  Identity Verified
Canada
Local time: 15:12
Member (2008)
French to English
+ ...
How many variables do translators control? Aug 25, 2016

Spence Green wrote:

The single variable that translators can control is hourly productivity and thus the *effective* hourly rate.


Er, no. Translators can control much more than that single variable. Such as who they will accept or not accept as a client, what type of document they will specialise in, what price they will charge, what effort they will put in to marketing their services and expanding their client base, and the list goes on. Hourly productivity is only one of a host of variables translators control, and not even necessarily the most significant one affecting their income.


 
Tom in London
Tom in London
United Kingdom
Local time: 19:12
Member (2008)
Italian to English
Who? Aug 26, 2016

Spence Green wrote:

... we're excited ...


Calm down, calm down. Who's "we"?


 
Paula Rennie (X)
Paula Rennie (X)  Identity Verified
Canada
Local time: 15:12
French to English
"But another rational strategy is to gain market share by reducing the price." Aug 26, 2016

The only market share you gain by lowering your price is at the bum-end. That would be the irrational thing to do.

 
Spence Green
Spence Green  Identity Verified
United States
Local time: 12:12
Member (2014)
English to Arabic
+ ...
TOPIC STARTER
Re: How many variables do translators control? Aug 26, 2016

John Fossey wrote:

Such as who they will accept or not accept as a client, what type of document they will specialise in, what price they will charge, what effort they will put in to marketing their services and expanding their client base, and the list goes on. Hourly productivity is only one of a host of variables translators control, and not even necessarily the most significant one affecting their income.


It's true that there are other variables (see my earlier response), and that my initial post was imprecise.

That said, complete discretion in choice of clients would seem to indicate a supply shortage in the market. Great translators are always in-demand, but it seems that many people have less flexibility. Otherwise, why would any translator accept a basic post-editing job with a flat rate discount? It seems like everyone would reject this work in favor of higher paying clients.

Document specialization is again affected by market conditions. You could choose a specialization for which there is little demand, or for which demand dries up. Or you could choose a specialization that is lucrative, but other competitors appear. It's very difficult to control these outcomes.

Your personal translation performance---both the quality of your translations and the speed at which you produce them---are factors that you alone control. I should have added marketing performance to my first post. How you market your translation skills is obviously important and under your control.

My observation is simply that the per-word price is discussed frequently, but that the throughput parameter, which also affects hourly rate, is less frequently discussed.


 
Spence Green
Spence Green  Identity Verified
United States
Local time: 12:12
Member (2014)
English to Arabic
+ ...
TOPIC STARTER
Re: Who? Aug 26, 2016

Tom in London wrote:

Calm down, calm down. Who's "we"?


A link is provided to an article that should resolve the pronoun.


 
Spence Green
Spence Green  Identity Verified
United States
Local time: 12:12
Member (2014)
English to Arabic
+ ...
TOPIC STARTER
Re: "But another rational strategy is to gain market share by reducing the price." Aug 26, 2016

Paula Rennie wrote:
The only market share you gain by lowering your price is at the bum-end. That would be the irrational thing to do.


Pricing can be used to achieve various business outcomes, market share expansion being one of them.

Quick read:
"A business can use a variety of pricing strategies when selling a product or service. The price can be set to maximize profitability for each unit sold or from the market overall. It can be used to defend an existing market from new entrants, to increase market share within a market or to enter a new market." https://en.wikipedia.org/wiki/Pricing_strategies

Longer read:
Michael Porter. Competitive Strategy. https://www.amazon.com/Competitive-Strategy-Techniques-Industries-Competitors/dp/0684841487


 
Spence Green
Spence Green  Identity Verified
United States
Local time: 12:12
Member (2014)
English to Arabic
+ ...
TOPIC STARTER
Re: The MT postediting ISO standard Aug 26, 2016

Jeff Allen wrote:

As for MT, it is not so simple. To be able to use a statistics-based MT system, it requires having much content upfront to train the system. This is not always possible for each customer, and especially for each language.
On the other hand, rule-based systems are quite good for small projects of a 5000 - 500,000 words to quickly analyze the translation project, extract out all terminology, quickly create a customer dictionary with variants, and apply it immediately in combination with the TMs. It is possible to do all of this, including on-the-fly changes in real-time post-editing mode.



Hi Jeff:

This is one of the core problems solved by online training. An adaptive statistical system typically starts from a baseline model trained on a considerable quantity of background parallel data (e.g., 160M sentence pairs for en-fr, in our case). Then each new segment that is either confirmed by the translator or uploaded from TM is added to the baseline for domain adaptation. In our system, each update takes less than 500ms; the system learns at a rate of about 10k segments / hour.

Batch-trained statistical systems require full re-training to add a single segment to the model; adaptive / online systems don't.

As for RBMT, recent WMT shared tasks have shown inferior performance (according to a blind human quality rating) relative to statistical and neural systems. In 2016, just one RBMT system was entered into the evaluation campaign. It lagged the other systems considerably for en-de and ru-en, although it won en-ru. 2014 was the last year in which rule-based systems were included as baselines, and they lagged the other systems by wide margins.

It remains true that for low-resource language pairs one can resort to rules. But that time could also be profitably spent collecting data, as Microsoft did for White Hmong.


 
Pages in topic:   < [1 2]


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


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

Adaptive machine translation is coming? Or it's already here?






CafeTran Espresso
You've never met a CAT tool this clever!

Translate faster & easier, using a sophisticated CAT tool built by a translator / developer. Accept jobs from clients who use Trados, MemoQ, Wordfast & major CAT tools. Download and start using CafeTran Espresso -- for free

Buy now! »
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 »