tag:blogger.com,1999:blog-2998896801371077929.post4214498732134013201..comments2023-05-30T10:07:44.851+02:00Comments on Language mystery: Would I advise my grandchildren to translate?Victor Dewsberyhttp://www.blogger.com/profile/18342577630994069368noreply@blogger.comBlogger7125tag:blogger.com,1999:blog-2998896801371077929.post-43337350875574188682012-05-16T13:14:23.876+02:002012-05-16T13:14:23.876+02:00Thanks for commenting, Krzysztof, but if you look ...Thanks for commenting, Krzysztof, but if you look carefully I think you will find that I did not say what you think I said. I did not actually comment on quality "tending to be lower" in PEMT compared with human translation, I merely pointed out one of the potential problems with PEMT and commented that with my own working style, human translation is more reliable.<br /><br />On the other hand, your comment offers an interesting example of the "have your cake and eat it" approach that can sometimes be seen in MT advocates. It is often argued that for many MT texts post-editors should not be so fussy about "quality", and should focus on ensuring that the text is fit for purpose, or in many cases merely "good enough", and translators who advocate translation expertise are branded as "quality wimps". To contrast with this, it is then suggested (e.g. in your comment) that "goot MT" is often better than human translators. So the practising language experts lose both ways - either we are too focused on quality, or we should give up because MT beats us hands down. After all, MT gets numerals and dates right (wow, ain't that revolutionary!!!!).<br /><br />In reality, the issues on both sides are far more complex. We all know that there are negligent human translations in which details such as numerals and dates are mistyped and the written style is poor. Sometimes the raw MT output is better than a poor human translation, at least in part. And many expert translators integrate IT tools such as TM programs or even MT into their workflow. But MT is dependent on human experts, and for most purposes a good human translation will be significantly more reliable, readable and consistent.Victor Dewsberyhttps://www.blogger.com/profile/18342577630994069368noreply@blogger.comtag:blogger.com,1999:blog-2998896801371077929.post-8961473061834926182012-05-14T10:19:17.850+02:002012-05-14T10:19:17.850+02:00@Victor, I don't agree that the quality of pos...@Victor, I don't agree that the quality of post-edited MT output tends to be lower compared to human translation. There are issues where goot MT is actually better than humans (like numerals or dates). But MT post-editors have to pay attention to other kinds of errors, more likely to be found in machinely-translated text. It also usually requires more editing work than (good) human translation.Krzysztof PrzyĆuckihttp://www.btinfo.plnoreply@blogger.comtag:blogger.com,1999:blog-2998896801371077929.post-68758116216175828612012-03-13T10:45:50.793+01:002012-03-13T10:45:50.793+01:00@Kirti: thanks for commenting - I feel honoured. A...@Kirti: thanks for commenting - I feel honoured. And thanks for saying "There is more to humans than data can capture and predict." And thanks, too, for saying that you would not be concerned about the future if you were a translator. The perception of many freelancers is that MT advocates aim to do away with us (which is no wonder when you see the rhetoric of TAUS and others), so it is helpful for you to formulate so clearly that you expect us to be firmly in the game for many decades to come. These two quotes are highly tweetable (give me a couple of minutes to put them up there).<br /><br />It is interesting to note the major role which you see for MT, i.e.: "to accelerate the production of business content that is predictable (e.g. most user documentation)". To some extent I see your point, but this can quickly lead into dangerous water, especially where such texts include safety-critical content (e.g. medical, electrical or mechanical topics). I would trust my ability to avoid such pitfalls when I am translating safety-critical texts, but post-editing such texts after they have passed through MT (especially domain-specific trained MT systems) would requires an incomparably higher level of detective work. In short, I would trust a text that I have translated, but I would not trust a text where I have post-edited MT output.Victor Dewsberyhttps://www.blogger.com/profile/18342577630994069368noreply@blogger.comtag:blogger.com,1999:blog-2998896801371077929.post-30564191692210672302012-03-12T18:46:40.858+01:002012-03-12T18:46:40.858+01:00For the record, while I see MT making great stride...For the record, while I see MT making great strides and enabling much more content to be translated at some level of "acceptable quality" , I do not subscribe to the view that by 2020 everything will be free and all words to be translated will have been seen before. I am doubtful that this could even happen in 100 years<br /><br />I also don't buy the singularity theories. There is more to humans than data can capture and predict.<br /><br />We have already seen at Google that hypothetically adding 5 billion new words of TM does not improve their Spanish engine since the technology has already learned all it can or more likely the data has too much noise in it for this to occur.<br /><br />While the possibility for very high quality MT for high quality business source materials exists, much of what is thrown into MT is of dubious quality in the first place. I am skeptical that this will change and I expect that "good" translators will always be rare and special and that whenever it really matters to get it right, human translators will be needed and necessary to ensure the MT is accurate and useful.<br /><br />MT will be great to accelerate the production of business content that is predictable (e.g. most user documentation)and that has a short life and possibly high value for a point in time and then disappears. <br /><br />But in terms of communication, even for social networks we are seeing that MT is a long way from providing real value in Twitter and FB. <br /><br />IMO MT will provide value, and accelerate translation efficiency, one domain at a time and will require translator guidance for my life time at least. <br /><br />I would not be concerned if I were a translator as many translation knowledge related challenges I expect will continue until we reach the Star Trek age and even then will we be really able to translate Shakespeare or Chaucer or Tagore?Kirti Vasheehttps://www.blogger.com/profile/16795076802721564830noreply@blogger.comtag:blogger.com,1999:blog-2998896801371077929.post-88214776380181341892012-03-02T18:02:00.159+01:002012-03-02T18:02:00.159+01:00@Steven: Actually, I took a liberty there. Hidden ...@Steven: Actually, I took a liberty there. Hidden away at the bottom of the "About" page is the name of the society when it was founded in 2004: "Translation Automaton User Society". I assumed the missing "i" was to throw intruding spybots off the scent (because TAUS knows about that sort of thing), so I gave them an extra "i" free of charge.<br />@Jordi: What a let-down - I thought they had actually invented the automatic blog-article-writing robot .Victor Dewsberyhttps://www.blogger.com/profile/18342577630994069368noreply@blogger.comtag:blogger.com,1999:blog-2998896801371077929.post-74291752451545444432012-03-02T16:53:32.268+01:002012-03-02T16:53:32.268+01:00Click on "Perspectives" at the top of th...Click on "Perspectives" at the top of the article and you will see that this article was penned by Jaap, TAUS' boss and chief futurist. <br /><br />Once futurists were called sci-fi writers. And before that, fantasy writers. Is he right? I don't know, but I hope not so.Jordi Balcellshttps://www.blogger.com/profile/00200189720838442406noreply@blogger.comtag:blogger.com,1999:blog-2998896801371077929.post-12039140110346826502012-03-02T15:50:06.457+01:002012-03-02T15:50:06.457+01:00Ah, so THAT's what TAUS stands for. It's o...Ah, so THAT's what TAUS stands for. It's one of several privacy-violating plug-ins I keep disabled on the CAT software I use.Steven Capsutonoreply@blogger.com