Topic: | Re:Re:Re:Epistemology: questions regarding the discrete model |
Posted by: | Patrick E.C. Merlevede, MSc. (jobEQ.com) |
Date/Time: | 04/01/2003 10:00:21 |
Message to all: The website "nl" pointed to is a goldmine! If you access it from it's main page: http://cogprints.ecs.soton.ac.uk/ You'll find many other interesting articles there, such as "Computing Machinery and Intelligence" by Turing, Miller's magical number 7 Plus minus 2, etc. While, unfortunately for those researching NLP's roots, most of the 1551 research papers published here are more recent (1980 or later), it shows how the field of cognitive science had developed, with all its subfields, of which NLP could have been one, if it would have done some efforts to build rapport with the scientific worldview (especially with the rules of doing research and publishing about it). What one will immediately note is that most entries in the Cogprints archive are "refereed research literature". Because of the "peer review system", even if that doesn't work perfectly, cognitive scientists will consider that these entries have some "credibility", which is precisely NLP's biggest problem. IMHO, on pages 350-354 of "Whispering", the authors acknowledge that problem by addressing one of the most important aspects of the current NLP community: a total lack of peer review (even if I don't believe Carmen & John's recommendations will be realized soon). The reason the mechanism of research works in other fields is that those researchers are paid by universities and by research grants to do research, and need to have peer reviewed publications in stead of keeping their job. Of course only few NLPers are paid for the modeling they do, and few have to care about their research being published, let alone being peer reviewed. Publishing an article in an NLP magazine is more an act of idealism or maybe serves as one’s personal marketing, and as I know from personal experience, once you have achieved a certain level of credibility, no peer review what so ever takes place, on the contrary, publishers are happy you submit an article and publish it without any form of criticism. Any only on a few occasions published articles are publicly criticized in other articles. (In other scientific fields, writing a critical review of another article is much more motivating, since, when your criticism gets published, that gives you some other "points" which are helpful to keep your job or get some more research funding.) In the field of NLP, researchers consider it a hassle to respond to criticism: it only takes up their time and competes with “real worked” (work that generates money, that is, such as doing training). For me, the biggest historical mistake made around 1978 is setting up DOTAR as a commercial company outside the university, in stead of lobbying to create a research institute within the framework of the University of California. Patrick www.merlevede.biz |