| Pages in topic: < [1 2 3 4] > | Machine translation: your experience with the various MT programmes? ("state of play") Thread poster: Barnaby Capel-Dunn
|
|---|
Vadim Pogulyaev 
Thailand
Local time: 18:54
Member (2007)
English to Russian
Please correct me, if I am wrong. You would prefere proofreading of machine translation (dumb and incomprehensible) to translation from scratch (making your own product). You want to make living by converting unreadable sets of words, produced by MT, into "satisfactory", "fit-to-purpose" text. And you'll still enjoy your job.
It's upto you. Good luck. In a good hour.
There are reasons why neither statistical, nor algorythmic MT will never work for translation into slavic langu... See more Please correct me, if I am wrong. You would prefere proofreading of machine translation (dumb and incomprehensible) to translation from scratch (making your own product). You want to make living by converting unreadable sets of words, produced by MT, into "satisfactory", "fit-to-purpose" text. And you'll still enjoy your job.
It's upto you. Good luck. In a good hour.
There are reasons why neither statistical, nor algorythmic MT will never work for translation into slavic languages. So I am not in your boat. ▲ Collapse
| | | | gianfranco
Brazil
Local time: 08:54
Member (2001)
English to Italian
+ ...
| See you in 10 years from today | Mar 24, 2008 |
Vadim Poguliaev wrote:
...
Please correct me, if I am wrong. You would prefere proofreading of machine translation (dumb and incomprehensible) to translation from scratch (making your own product). You want to make living by converting unreadable sets of words, produced by MT, into "satisfactory", "fit-to-purpose" text. And you'll still enjoy your job.
It's upto you. Good luck. In a good hour.
Yes, I don't mind if I work translating a document, reviewing some documents translated by another human or post-editing some machine generated text. A revision is a revision.
I also don't mind if I work with pen and paper or a computer, with or without a CAT tool, with a paper or an electronic dictionary, etc.
I just said that I have no objection in pronciple, even if some of the tasks could be more interesting than others, it is often a matter of habit and we can get job satisfaction from many different things, including some apparently boring tasks.
BTW, translation can be very often stupefyingly boring, and yet we do it and occasionally we even enjoy it. Go figure...
Vadim Poguliaev wrote:
There are reasons why neither statistical, nor algorythmic MT will never work for translation into slavic languages. So I am not in your boat.
Never say never. See you in 10 years from today...
bye
Gianfranco
| | | | | If it happens, what will the consequences be? | Mar 25, 2008 |
This is getting interesting. If we follow Gianfranco's logic (and he is not alone in having that logic), then indeed, he is probably right in his forecast. However, if this really is to occur, then I would be interested to know what you all think the consequences of that will be. I see the following possibilities:
- The per word rates for translation in general will be smashed. You think rates are low now? Ain't seen nothin' yet! On the grounds that we are only editing/proofreading ... See more This is getting interesting. If we follow Gianfranco's logic (and he is not alone in having that logic), then indeed, he is probably right in his forecast. However, if this really is to occur, then I would be interested to know what you all think the consequences of that will be. I see the following possibilities:
- The per word rates for translation in general will be smashed. You think rates are low now? Ain't seen nothin' yet! On the grounds that we are only editing/proofreading and not translating, we will probably be expected to charge less than for traditional translation. Newbie translators and seasoned translators alike will probably have lots of trouble estimating the time it will take to finish the job, causing them to lose even more money.
- Because of the unpredictabilities associated with editing/proofing MT translations, people will charge less and less by the word and more and more by the hour. Who knows - maybe this will eventually drive rates up, because people will be paid for their time and not for their productivity.
- As Gianfranco says, there will always be room for human translation. BUT! If most translation jobs are done using MT, isn't it logical to expect that people will think that MT can entirely replace humans, and that even jobs that still should be done by humans will be processed using MT? This could lead to some very unpleasant surprises, especially in technical and legal translation. Then again, maybe after a few of those very unpleasant surprises, people will start appreciating human translators more, which would balance things out and eventually even contribute to rate increases for us humans. Yay!
Any thoughts? Or any possibilities I haven't though of? Or even both? ▲ Collapse
| | | | Samuel Murray
Netherlands
Local time: 13:54
Member (2006)
English to Afrikaans
+ ...
| There is something to be said for it | Mar 25, 2008 |
gianfranco wrote:
Yes, I don't mind if I work translating a document, reviewing some documents translated by another human or post-editing some machine generated text. A revision is a revision.
There is something in what you say. In fact, since machines typically make certain mistakes it may be easier to see some of the mistakes because they are so predictable. It may even be an idea to give such documents two revisions -- one that marks potential typical MT errors, and one that is the normal, careful revision you'd give any human written text anyway.
On the other hand, revising an MT job could be like trying to spellcheck a document that has been spell-checked by a mind-less human who simply accepted all suggestions from the spell-checker. In a sense, it is more difficult because all the words in the document are words that are spelt correctly.
I suspect some revisers tend to find only major errors -- errors that jump at them from the page, and since an MT translation is likely to contain mostly grammatically correct phrases and sentences, fewer errors jump out from the page (and more errors remain hidden, if you revise too fast). But that is a problem with the revisers, not with the translators (whether hardware, software or wetware).
| | |
|
|
|
Samuel Murray
Netherlands
Local time: 13:54
Member (2006)
English to Afrikaans
+ ...
| You can just revert to earlier methods | Mar 25, 2008 |
Viktoria Gimbe wrote:
On the grounds that we are only editing/proofreading and not translating, we will probably be expected to charge less than for traditional translation.
Not really. You charge per hour at the moment, and you'll be charging per hour ten years from now. The only difference is that your per-time to per-unit conversion will be done differently.
Newbie translators and seasoned translators alike will probably have lots of trouble estimating the time it will take to finish the job...
How do you currently estimate how long a proofread/revision/editing will take?
I suspect it may actually be somewhat easier to determine the duration if you know which MT system was used, because you would use semi-automated tools to mark typical problems that occur with such and such MT system.
If most translation jobs are done using MT, isn't it logical to expect that people will think that MT can entirely replace humans, and that even jobs that still should be done by humans will be processed using MT?
Take for example what happened when DTP became available for everyone (not just DTP people). Now you get Microsoft Publisher, and even word processors can do text and graphics layout. This means more people print brochures and newsletters, because it has become more accessible to them. Did lots of typesetters and DTP folks lose their jobs? Did they become more specialised at what they do? Not really -- the real customers of DTP people are still businesses who realise that you can't publish the brochure of a multi-million dollar project using MS Publisher and that you'd be better off paying someone who's using Corel Draw or some of those programs (with the added skills to do proper DTP anyway).
I'm sure you'll get more "translation" service providers who simply do MT and revise the output, and real translators will have the job to educate clients about the dangers of using a service provider who is not a real translator but merely an MT-output reviser.
| | | |
Victoria Gimbe wrote:
This is getting interesting. If we follow Gianfranco's logic (and he is not alone in having that logic), then indeed, he is probably right in his forecast. However, if this really is to occur, then I would be interested to know what you all think the consequences of that will be.
Based on the premise that MTs are robotic machine tools producing serial items, that makes handcraft items more expensive. Therefore, the future seems greatly rewarding for human translators with or without MTs, CATs, DTPs, etc.
To ensure that a machine-generated translation will be of publishable quality and useful to a human, it must be reviewed and edited by a human.
Finally: what is the use of using MTs? Time. Time saved per word when translating. Time gained per hour when plainly editing.
I think there is nothing to fear, and nothing to criticize or censure either. Worry about delivering a good translation, and keep your clients smiling.
| | | |
Thanks so much everyone for your illuminating comments and contributions. I personally have learnt a lot from them.
It just struck me that, whatever else MT might be able to do over the years, it will NEVER be able to handle the sort of language used in this thread. So maybe that's the way ahead for us translators? Back to those skills in rendering relaxed, conversational, idiomatic, literary and academic texts that drew many of us to the profession in the first place!
| | | | | A word from Google | Mar 26, 2008 |
Matt Glotzbach, Google's director of product management for Enterprise:
"What you need for real-time automated machine translation is large amounts of compute power, which we have, and large amounts of data, which we have," said Glotzbach. "Imagine a system that can do on-the-fly translation of things like e-mail, documents and IM chat. That's a feature of [Google Apps] you can see on the horizon."
| | |
|
|
|
Samuel Murray
Netherlands
Local time: 13:54
Member (2006)
English to Afrikaans
+ ...
Barnaby Capel-Dunn wrote:
Matt Glotzbach, Google's director of product management for Enterprise:
"What you need for real-time automated machine translation is large amounts of compute power, ... and large amounts of data..." said Glotzbach.
Okay, so you need large amounts of compute power and large amounts of data, and then you'll have on-the-fly MT. So, who's got large amounts of compute power and large amounts of data? Google of course. And does Google have on-the-fly MT yet? Noooooo... so it's just vapourware.
| | | | Jeff Allen
France
Local time: 13:54
Multiplelanguages
+ ...
| some MT state-of-the-art case studies | Nov 11, 2008 |
Barnaby,
Thanks for starting this thread. Sorry I've seen it late.
I'll have to give my reply in a few instalments because there are lot of embedded issues in the comments made in this thread.
1) Online MT portals versus professional desktop and enterprise level MT software/systems.
This has already been mentioned by Carvallo in this thread, and which I have said in many posts on this site as well.
The online portals simply create a larg... See more Barnaby,
Thanks for starting this thread. Sorry I've seen it late.
I'll have to give my reply in a few instalments because there are lot of embedded issues in the comments made in this thread.
1) Online MT portals versus professional desktop and enterprise level MT software/systems.
This has already been mentioned by Carvallo in this thread, and which I have said in many posts on this site as well.
The online portals simply create a large of postediting time. Use the professional and expert versions of the desktop or enterprise level MT systems. These are those which contain the features to perform the tasks which reduce the postediting time at the end.
The PROMT and SYSTRAN systems are regularly evolving with many new features. I've used both extensively, and have used/tested a variety of other MT systems (not just online portals, but the desktop and enterprise level systems) over the years.
I've just recently the MT postediting site (http://www.geocities.com/mtpostediting/) which contains all of my project case studies, productivity studies, walkthroughs, etc on using MT systems from the perspective of a translator.
These articles provide the workflow processes, and provide the results. The more recent ones give more productivity statistics.
Other topics that I want to cover in subsequent posts in this thread are the following:
2) Content types favorable for MT
3) The enterprise offer statistical MT approach which includes training on TM databases
etc
4) the mixing of TM and MT technologies
5) cost model shifts over time
Jeff
[Edited at 2008-11-11 16:10] ▲ Collapse
| | | | Jeff Allen
France
Local time: 13:54
Multiplelanguages
+ ...
| MT quality depends on source content | Nov 13, 2008 |
And now for topic 2
2) Content types favorable for MT
Is it possible to use MT for various topics. Yes, I've given a list of those that I've used it for at the following post:
Machine Translation: yes used for professional translation work
http://www.proz.com/forum/translator_resources/6526-machine_translation_do_you_use_it.html#177427
* call for tenders
Viktoria Gimbe wrote:
Recently, one of my regular clients asked me to evaluate a project where MT was to be used to help their English-speaking clients understand and answer questions in a call for tenders written in French. ...
They wanted to use MT to translate the requirements section into their own language in order to be able to answer the associated question. ...
Google's MT engine was chosen for this purpose and part of my evaluating the project was to take some paragraphs from the call for tenders and process them with Google to try to predict what portion of the French text the client would need assistance with. Let me tell you, I was amazed with the results. No matter what paragraph I took, the output was a pretty accurate rendering of the source concept. Of course, the wording wasn't always perfect and if any of it had to be published, it all would have required editing..
This experience with Google taught me one thing: some texts are naturally adapted for MT and some aren't. The subject was a call for tender and there weren't many technical terms in it. Moreover, the source text was of excellent quality, free of any ambiguity, the style was bland and there were no cultural references.
As Viktoria has indicated, RFPs (and RFIs, RFQs, etc) are usually well written. I would say that they are rehashed and reviewed to death. But the simple reason for that is that they are usually time-constrained for the reply, and those who issue them do not want the respondents calling up every 15 min to enquire about every 3rd point. The objective is to issue a document that is both comprehensive and comprehensible and to receive a standardized set of responses of which to evaluate each proposal.
As for technical vocabulary, it really depends on the topic and the organization. I've worked with very generic RFPs which serve a wide range of possible response topics. And I've also written and responded to very technical RFPs. As noted in my link above, RFPs (RFIs) have been types of documents that I have used MT for, and this includes the adapted translation of an RFI for publication, for which the MT postedited version was drafted in record time, and was issued and received replies which led to very large contracts. This MT postediting project is discussed in one of my conference papers, mentioned in the following post.
MT user case study showing that MT is not useless
http://www.proz.com/forum/money_matters/25140-minimum_rate_just_like_ebaycom_has_to_be_established-page4.html#174679
* Newswire and news stories are another type of content which usually obtains quite decent MT results. The reason for this is that MT vendors focus a lot on creating systms that can do good content gisting, and news is a content source that is consumed multilingually. MT vendors often use online news articles as a way to sell the product, asking the customer prospect to pick any news story from the same day news and to MT process it online. I've used that approach myself, and it has worked. The translation teams that I led at Carnegie Mellon spend a couple of years translating newswire stories, and we measured their productivity in creating the translated text, and recycling it in a translation memory approach. There were normally 18-20 topic types, and I've recent investigated with another news organization that has 17 general topics. Our findings in translating several languages on the entire range of topics were that the sports topic and the finance topic were the most difficult for the human translators to work on, especially since we kept filtering out content that they had already seen, and which had a minimal threshold of already viewed vocabulary. This resulted in pushing for sentences which constantly had new vocabulary, and which was thus more difficult to translate. That case study is in the following paper.
EBMT doesn't necessarily take several person years to build up
http://www.proz.com/forum/translator_resources/37742-is_this_the_future_automatic_simultaneous_translation_within_5_years-page2.html#270123
which contains the following paper:
ALLEN, Jeffrey and Christopher HOGAN. 1998. Expanding lexical coverage of parallel corpora for the Example-Based Machine Translation approach. In Proceedings of the First International Language Resources and Evaluation Conference (LREC98), 28-30 May 1998, Granada, Spain. Vol. 2, pp. 747-754.
http://www-2.cs.cmu.edu/afs/cs.cmu.edu/user/chogan/Web/lre98.zip
* content which follows writing guidelines
Viktoria Gimbe wrote:
If MT doesn't provide translations of acceptable quality, it isn't necessarily because MT is not smart. Some texts don't require much intelligence to understand and then render in a different language. What stands in the way of MT success, in my opinion, is poor source texts, more specifically texts that were not written with translation in mind. MT aside, I often have to work with texts that clearly weren't ready for translation and not even for publishing. Too many sentences with words used in the wrong context (sorry no compute) and incomplete sentences (no verb, no subject, etc.). Even for human translators, this can cause a lot of problems.
Yes, it's the garbage in / garbage out (GIGO) principle. Content that follows style guides, international communication writing guidelines, or whatever you want to call them, will produce better translatable content, whether it be translated by human or machine. The extreme of that is controlled language writing for which a good number of such writing systems and guidelines have been designed with MT in mind. See:
http://www.geocities.com/controlledlanguage/
Some translations providers also refer to this as source content optimization and have methods for making the content easier to translate/localize.
* terminology work
Any project for which terminology work is done upfront and/or during the project usually produces better quality translation. All projects for which I have provided custom terminology components (including high quality human translated branded content) have led to repeat customers. The same is true when terminology work can be entered into MT dictionaries and is quickly recycled back into the MT system for all subsequent translations. This brings significant improvement for the MT results. Several of my case studies on MT dictionary building at the MT postediting site (http://www.geocities.com/mtpostediting ) describe this in detail.
I believe that MT can be used in nearly every subject area, but it requires indepth technical knowledge on the subject matter, and a good mastery of MT dictionary principles in order to avoid overengineering the dictionary, and to avoid degrading translations for other possible fuzzy match contexts.
Jeff
| | | | | Sole MT versus Convergence | Nov 22, 2008 |
The question of MT as a sole approach is actually not the case...
MT can only be one component within a greater convergence of numerous approaches.
Technological progress Web 2.0 is driven by Convergence and Mashups nowadays.
http://en.wikipedia.org/wiki/Convergence#Computing_and_technology<... See more The question of MT as a sole approach is actually not the case...
MT can only be one component within a greater convergence of numerous approaches.
Technological progress Web 2.0 is driven by Convergence and Mashups nowadays.
http://en.wikipedia.org/wiki/Convergence#Computing_and_technology
http://en.wikipedia.org/wiki/Mashup_(web_application_hybrid)
The "state of play" as convergence is imminent with the emerging of
Google Translation Center.
Evaluations performed by the National Institute of Standards and Technology NIST Open Machine Translation (MT) are performed annually since 2001. Based on the Bilingual Evaluation Understudy method for rating translation accuracy, Google scored first place in a 2005 evaluation by the National Institute of Standards and Technology evaluation.
http://www.nist.gov/speech/tests/mt/
http://en.wikipedia.org/wiki/Bilingual_evaluation_understudy
http://www.nist.gov/speech/tests/mt/2005/doc/mt05eval_official_results_release_20050801_v3.html
Beyond this, Social Media, Wikis and Blogs are already NOW delivering significant strategic impact to external client structures and interfaces, which then reach into external and internal translation processes via Collaborative Innovation. So there is no need to twiddle your thumbs and wait for Google Translation Center.
Within Google's software as a service business model, a new structural Convergence approach is emerging:
Statistical Machine Translation + Crowdsourcing + Web 2.0 Translation Memory
http://en.wikipedia.org/wiki/Software_as_a_service
http://en.wikipedia.org/wiki/Crowdsourcing
Yet as a mechanical system, it needs to be pointed that Google Translation Center does not organize Collaborative Translation, which is organistic.
http://collaborative-translation.ning.com/group/googletranslationcenter
Besides, Wordfast and VLTM, together with Transcription, already enhances the mix to the following Web 2.0 convergence opportunity:
Statistical Machine Translation + Crowdsourcing + Web 2.0 Translation Memory + Audio Transcription
http://www.wordfast.net/index.php?whichpage=jobs ▲ Collapse
| | |
|
|
|
| MT in which language? | Nov 24, 2008 |
Seems to me that most of you are referring to translation between the large World Languages, English, French, Spanish, and German in this thread, but what about the smaller languages? From my own region of the world, I can name Swedish, Danish, Norwegian, Sami, Finnish, and Finnish-Swedish, and I say that we're not facing any imminent 'danger' from MT in our neck of the woods...
I mean, the market and possible ROI is just too small for anybody to invest in creating MT for these lang... See more Seems to me that most of you are referring to translation between the large World Languages, English, French, Spanish, and German in this thread, but what about the smaller languages? From my own region of the world, I can name Swedish, Danish, Norwegian, Sami, Finnish, and Finnish-Swedish, and I say that we're not facing any imminent 'danger' from MT in our neck of the woods...
I mean, the market and possible ROI is just too small for anybody to invest in creating MT for these languages, together with some thousands of smaller languages on this planet.
We've seen it with several efforts on developing Speech Recognition in Norwegian. Over the last 10 years, several enterprises have broken their back on such an effort, leaving us with no Speech Recognition applications for Norwegian today.
As I'm sure most of you know, the work and investment required to create MT for a language is not dependent on the size of the language, nor the market prospect, so Norwegian MT will be just as costly as that for English or Spanish, or Chinese/Japanese for that matter.
Accordingly, I'd say that translators in small languages may lower their shoulders. And, a fully functional MT for the World Languages, not requiring human post-editing, will not be on the market for another 10 years, at least!
I may be wrong, of course, but while waiting, I'll be happily translating into/from Norwegian and sleep well at night!
Best,
Roald Toskedal ▲ Collapse
| | | | Jeff Allen
France
Local time: 13:54
Multiplelanguages
+ ...
| statistical MT approach + TMs | Nov 24, 2008 |
3) The enterprise offer statistical MT approach which includes training on TM databases
Up until recently, the reference to "MT" was basically with regard to Rule-Based MT (RBMT) systems.
Statistics-based MT (SMT or SBMT) has been in the background for many years, waiting to emerge. It did so first with Language Weaver as a commercial system. Asia Online is another commercial offer.
Google also within the past year or so slowly started transitioning by phasing ... See more 3) The enterprise offer statistical MT approach which includes training on TM databases
Up until recently, the reference to "MT" was basically with regard to Rule-Based MT (RBMT) systems.
Statistics-based MT (SMT or SBMT) has been in the background for many years, waiting to emerge. It did so first with Language Weaver as a commercial system. Asia Online is another commercial offer.
Google also within the past year or so slowly started transitioning by phasing in their own home-grown SBMT engine to replace the SYSTRAN system that they were using for many year. Just this year they have completely migrated over to their SBMT engine for all languages.
Commercial SBMT vendors state publicly that TMs (which is what has been called Example-based MT in the past) to train their systems. This fine-tunes the SBMT system on approved corporate content that has already been translated.
If you train an SBMT system with 80,000 Translation Units, the SBMT system is going to do significantly better than any traditional TM tool. The reason for this is that SBMT systems require massive amounts of processing power (on dozens of servers) to run the statistical analysis throughout all of the content in a parallel way, not running in a linear/sequential manner which TM tools do. This means that they are not just looking for matches of segments by setting a match threshold on a percentage, but rather that they are analyzing all occurrences of all words (actually rather then sequences of characters), and the distance between all the words so as to recognize and form "terms". I've give an example, the term "eat up" which I say to my kids probably 3 - 4 times a week at minimum. Most RBMT systems are restricted to recognizing "eat up" as in "eat up your food now". However, I say that phrase different ways "eat up your food", "eat your food up". An SBMT system is able to recognize all occurrences of these in a parallel TM database, and that that "eat" and "up" occur X number of different ways, and Y number of times each, across the TM, and within the new set of content to translate. It matches all of this up in parallel and proposes the statistically best translation proposal from the analysis of the source and target segments. Again, this is not a percentage based threshold setting but a fully statistical analysis.
Quite powerful, but then it is also a processing hog. It's not a machine learning method you can run on your home PC.
The downside of SBMT type systems has been that it was not possible to feed terminology databases into them (except the research type ones that I worked on), and there were no postediting interface modules. Both dictionary creation and postediting interfaces have been part of the RBMT systems for a few years. This has reduced the ease of plugging SBMT system into a workflow, but now the SBMT vendor companies are catching on and realize that there needs to be a more complete translation environment with the ability to do manual override with terminology databases, and also to make postediting easier to do.
I managed a translation laboratory of translation teams for 5 languages which built up parallel TMs for a few years. This was our EBMT-type data. We then took the SBMT processing method and extracted out candidates for "terms" purely based on the statistical analysis of all the content. The translators then reviewed the proposed terms and selected the best translation equivalents, which we fed back into the system to override the statistical analysis the next time around. Based on a year of data translated by each team (on a daily basis by 2 or more translators per team), we showed from statistical analysis the increase of the terminology (translated) which could be found when filtering out existing translated material, and only giving to the translators new content which they had not seen before. This is specifically showing how we adapted the SBMT approach to content translated by human translators as a way to improve the SBMT system, but again in combination with the EBMT methods (ie, Translation Memories). All of this is described and with graphs in our paper. (This is the conference paper cited a couple of posts higher in this thread, which written by Chris Hogan and myself, 1998, LREC98).
The SBMT vendors today are using a similar approach (SBMT trained on TM data), and some are just now starting to mention that they can override with terminology lists. That LREC98 paper describes it with the results.
It was interesting to see the change in attitude of the teams when we implemented this content filtering process. The teams started saying that they were constantly getting different information to translate (not the same repetitive texts already seen previously), and that it was more challenging. And during those months I spend a lot more time in the lab explaining all kinds of things to them (how to play baseball, how to play golf, movie titles, educational and organization terms, etc), because these were the new types of content that they were receiving that had not been provided beforehand.
If the MT vendors can provide the SBMT + TM approach at a rate that translation agencies and freelancers can pay for, and allow them to statistically analyze all of their existing TMs and process new content, this could be quite powerful.
Now Roald is right about this being more applicable to Major languages, which was my critique in a magazine in 2000-2001 about the Statistical method at that point in time. I'll address the Major/Minority language question in another post.
Jeff ▲ Collapse
| | | | | Thanks, Jeff! | Nov 24, 2008 |
Thanks for your insightful comments!
Unfortunately, I don't have experience from any development of MT.
I just used common sense and my business experience as a potential user of such an application. It may very well be that my translating and business experience are too limited to see the full picture - that the translator business/market will change way beyond our current manual approach, but by then it will be time for me to retire anyway...
Like I said,... See more Thanks for your insightful comments!
Unfortunately, I don't have experience from any development of MT.
I just used common sense and my business experience as a potential user of such an application. It may very well be that my translating and business experience are too limited to see the full picture - that the translator business/market will change way beyond our current manual approach, but by then it will be time for me to retire anyway...
Like I said, I may be way off in my predictions regarding the development time span, but I have some experience from development projects in general, and it can be amazing to see how long it may take from the necessary technology is ready to make it possible, to when the product is actually being released on the market...
I mean, in 1998, several science and computer magazines made a lot of 'hooplah' around the optical computers soon to hit the (consumer) market.
Well, we haven't seen any yet, have we?
My basic gut feeling is that I'll trust my own experience over marketing lingo and CBS on new technology any day.
That's how the dinosaurs got eradicated as well, you say?
Well, this 'dinosaur' has seen one too many 'stop-the-press' news on technology to jump on the wagon just yet...
It goes without saying that our jobs and markets will be different 10 years from now, but I do have the dream that one day, outsourcers will recognize the value of quality over pennies and that the human factor will be appreciated in the translation business like it should be.
Roald ▲ Collapse
| | | | | 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 » Machine translation: your experience with the various MT programmes? ("state of play") | 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! » |
| | Anycount & Translation Office 3000 | Translation Office 3000
Translation Office 3000 is an advanced accounting tool for freelance translators and small agencies. TO3000 easily and seamlessly integrates with the business life of professional freelance translators.
More info » |
|
| | | | X Sign in to your ProZ.com account... | | | | | |
Login to reply/comment 
