tag:blogger.com,1999:blog-6748877443699290050.post3220537053797168479..comments2024-03-27T23:43:31.674-07:00Comments on eMpTy Pages: The Growing Momentum of Machine Translation in Life SciencesKirti Vasheehttp://www.blogger.com/profile/16795076802721564830noreply@blogger.comBlogger6125tag:blogger.com,1999:blog-6748877443699290050.post-10172994917512487202018-12-18T11:08:24.037-08:002018-12-18T11:08:24.037-08:00Victor, I did look at the reference I provided, an...Victor, I did look at the reference I provided, and my point is that clinical trials today involve data collection from all over the world that need to e understood quickly. This data is continuous and fast flowing and needs to be understood rapidly for it to have business value to the drug discovery process. <br /><br />I have never suggested that MT will take over the translation industry, but where volume, turnaround speed, and cost matter, MT will only increase in value and use. We are seeing this combo in many more scenarios across industries. Google just reported that they translate 30 trillion sentences a year, which means 500B to 750B words a day. Would this be happening if there was no value?<br /><br />There are an increasing number of use scenarios where MT is used without any post-editing involved e.g. clinical trial data summarization. However, even here practitioners who use human linguistic steering will get better results. So there is always a role for humans to enhance quality if it matters to take it beyond basic possibilities.<br /><br />For the record, I remain one of those who does not believe that MT will ever replace human translation. I do however see MT increasing its role in fast flowing business content translations where speedy response to issues highlighted in the content is a business survival issue. The best places to use MT thus far are still those where it is simply not possible to use human translation processes for speed, volume and cost reasons. The overlap of MT&HT is tiny and humans always win in the overlap space.Kirti Vasheehttps://www.blogger.com/profile/16795076802721564830noreply@blogger.comtag:blogger.com,1999:blog-6748877443699290050.post-59798747269229489322018-12-15T02:04:10.254-08:002018-12-15T02:04:10.254-08:00This discussion seems to fog the issues, it provid...This discussion seems to fog the issues, it provides a few generalisations on machine learning coupled with even more bland statements of blind faith in MT (or NMT). And Kirti, your assertion in response to "Anonim" that "the bibliography has a reference to NMT" takes the biscuit. Have you actually read the reference and noticed how bland, vague and meaningless it actually is? This seems to me to be a fallacy of much modern scientific thinking - if something is referenced in a learned paper, then it must be meaningful. I was disappointed to see you fall into this trap.<br />There is a bandwagon of superficial IT advocacy which claims that computers will take over the whole translation industry before long. Recently, for example, the German SPD politician Lars Klingbeil said in a TV talk show that in a few years, translation and interpreting will not exist as a service because it will all be done by computers. We all know that talking rubbish is a necessary qualification for a career as a politician, but his glib prophecy of doom for intelligent human language processing is merely the tip of the iceberg. <br />We all know that MT has made progress and can play a role in the translation process. We also know that human translators are perfectly capable of providing less than perfect solutions. But in a serious discussion (as against mere advertising froth), consideration must also be given to the limitations of machine translation and the social, psychological and economic problems associated with the much-touted panacea "post-editing of machine translation".Victor Dewsberyhttps://www.blogger.com/profile/18342577630994069368noreply@blogger.comtag:blogger.com,1999:blog-6748877443699290050.post-32583705464326631552018-12-05T10:31:06.385-08:002018-12-05T10:31:06.385-08:00As you seem unaware, NMT is only possible because ...As you seem unaware, NMT is only possible because deep learning techniques are applied to linguistic data and thus ML and MT are not so far apart.<br /><br />Also, the modern drug discovery process requires and uses a great deal of data consolidated from global sources that include research studies, patient reports, social media experiences, communications and language differences amongst the DD project teams which can often be global teams. <br /><br />No matter how amazing the ML techniques used on drug chemistry evaluation and development, they will never replace clinical trials. Predominantly 3 phases with successively larger groups of patients across the world, to ensure safety and effectiveness. Clinical trials are the precondition for FDA approval and today these trials are often done all over the world. There is an overview here https://rctom.hbs.org/submission/benevolentai-from-machine-learning-startup-towards-new-drug-producer/ and note that the bibliography has a reference to NMT as the volume and frequency of multilingual data in both the core research and the clinical trials is increasing with each passing day. MT can indeed expedite and accelerate this process even though you may not be aware of it. Kirti Vasheehttps://www.blogger.com/profile/16795076802721564830noreply@blogger.comtag:blogger.com,1999:blog-6748877443699290050.post-43354610439070904092018-11-30T19:19:36.679-08:002018-11-30T19:19:36.679-08:00Sorry, but drug discovery has nothing to do with M...Sorry, but drug discovery has nothing to do with MT. This is bullshit. And Royal Society wrote about machine learning rather than MT. So... this is very disappointing you are making such dirty marketing tricks on this blog.Anonimhttps://www.blogger.com/profile/10010586973907915754noreply@blogger.comtag:blogger.com,1999:blog-6748877443699290050.post-33140792460837221882018-11-15T12:28:58.840-08:002018-11-15T12:28:58.840-08:00Yes, this is true and I wrote about this in an ear...Yes, this is true and I wrote about this in an earlier postKirti Vasheehttps://www.blogger.com/profile/16795076802721564830noreply@blogger.comtag:blogger.com,1999:blog-6748877443699290050.post-79823941545269884232018-11-15T11:32:10.671-08:002018-11-15T11:32:10.671-08:00It seems you are working in SDL marketing these da...It seems you are working in SDL marketing these days. Is this true?Anonymousnoreply@blogger.com