RequestsFrom: Ronald N. Kostoff (ONR)Date: Fri, 10 Nov 2000 08:41:59 -0500 Subject: Discussion: Accelerating Text Mining Implementation There is a critical publication issue limiting the dissemination and implementation of text mining technology that I have never seen addressed. Your data mining Web site/ KDNuggets newsletter may be the appropriate forum to discuss this issue. ISSUE Science and technology (S&T) text mining technology is developed mainly by the information technology community. The eventual end user for S&T text mining technology should, in many cases, be the technical domain expert. How well does the text mining technology presently diffuse from the developer to the technical specialty user? In particular, how does present text mining technology diffusion limit text mining technology implementation, and how can the diffusion process be improved to accelerate implementation? BACKGROUND The focus of this discussion is literature-based text mining technology diffusion. 1) Information Technology Literature Literature-based diffusion potentially allows the value of combining text mining technology with traditional technical analysis to be demonstrated to the performers/ managers/ evaluators in the technical specialty communities. Presently, most text mining technology diffusion pathways are very indirect, especially for the text mining technology applications to the technical specialty literature. This is a major reason for minimal implementation of S&T text mining technology in the technical specialty communities. The text mining researchers and developers appropriately publish new methodologies in the information technology literature. Unfortunately, the text mining researchers/ developers/ appliers also publish technical specialty text mining applications almost exclusively in the information technology literature, including the Proceedings of information technology conferences. Effectively, this is preaching to the choir. I know of almost no technical specialists (Physical/ engineering/ life sciences) who read the information technology literature. They have a difficult enough time keeping up with the extensive literatures in their own technical specialties. Paradoxically, they are the people who would benefit most from the potential efficiencies and Technology Watch capabilities that S&T text mining could offer. Thus, the present information technology literature route is extremely inefficient for transmitting S&T text mining application information to the prime technical specialty end user. 2) Technical Specialty Literature For the most part, the technical specialty literature contains no text mining studies. The technical specialty journal editors overwhelmingly interpret their charters as the traditional theoretical/ experimental/ computational technical approaches to S&T. Most of these editors do not include hybridization of text mining/ information technology with the technical specialty as part of their charter. Thus, the transmission of S&T text mining applications technology through the existing traditional technical journals is practically non-existent. 3) Hybrid Information Technology-Technical Specialty Literature A few of the more progressive technical disciplines have initiated journals that combine information technology with the technical specialty. Journal of Chemical Information and Computer Science (Chemistry) and International Journal of Medical Informatics (Biomed) are good examples, and they are first rate journals. There may be some technical discipline specialists (as well as information technologists) that read these hybrid journals, since they address topics of interest within the technical discipline. I remain to be convinced that a substantial fraction of the technical specialty communities reads these hybrid journals, especially relative to the fraction that reads the traditional technical specialty journals. DISCUSSION I believe that the best way to disseminate the S&T text mining information, to eventually maximize its implementation in the technical specialty community, is the direct approach. Somehow, the hybrid information technology-technical specialty papers need to find their way into the traditional mainline technical specialty journals. For the past year, I have been attempting to publish our S&T text mining application studies in the traditional technical specialty journals, in parallel with maintaining text mining methodology publications in the information technology journals. In some technical speciality disciplines, that are more advanced technologically and have some progressive journal editors, publication in the specialty literature has been straightforward. In other disciplines, especially those where publication tradition predominates, publication in the specialty literature has been a struggle. YET IT IS IN THESE SPECIALTY DISCIPLINES AND THESE SPECIALTY JOURNALS WHERE THE NEED FOR TEXT MINING DIFFUSION IS THE GREATEST! THE READERS OF THESE JOURNALS ARE THE END USERS THAT WE NEED TO TARGET WITH TEXT MINING TECHNOLOGY. How do we accomplish this goal? RECOMMENDATIONS I believe that proactive steps by government and industry are required. Perhaps grants to some traditional technical specialty journals that would subsidize an information technology section in each journal issue would be one solution. There may be other incentives, and other solutions, and it would be valuable if your readers could suggest them. The data mining/ text mining community contains very knowledgeable people who have thought about information technology implementation for some time, and they might have valuable ideas on improving dissemination and implementation of text mining technology. Do you think this is a topic that would be worth addressing in your newsletter/ Web site? RNK http://www.dtic.mil/dtic/kostoff/index.html http://www.sciquest.com/cgi-bin/ncommerce3/ExecMacro/sci_kostoff2.d2w/report ?nav_banner=special&Tmstmp=25884 |
Copyright © 2000 KDnuggets. Subscribe to KDnuggets News!