|Volume 3: No. 23|
Robert Jacobson (email@example.com or cyberoid @u.washington.edu) is founder, president, and CEO of WORLDESIGN, Inc. (Seattle), using virtual-reality concepts to build enterprise information systems (EIS), geo-based information systems (GIS), training, and decision-support applications for corporations. Bob is also a research associate professor in UWashington's Dept. of Landscape Architecture and is a design consultant to clients in the US, Asia, and Europe. He was co-founder and associate director of UWashington's Human Interface Technology (HIT) Laboratory. He has also written two books and has worked with the California State Assembly on telecommunications policy. His PhD in urban planning is from UCLA, following a Fulbright scholarship to study telecommunications in Scandinavia. Bob now co-moderates the sci.virtual-worlds newsgroup, which runs about 45 messages per night. If you're not on Usenet or NetNews, you can get a mail feed from Prof. Greg Newby (gbnewby @uxh.cso.uiuc.edu).
David Lewis is compiling a list of NLP discussion lists and resource directories. His first pass includes NL-KR, IRLIST, LINGUIST, EMPIRICISTS, CORPORA-LIST, LN, HUMANIST, TCC, and a list maintained in /pub/catalog on clr.nmsu.edu. Suggestions are welcome. David is also compiling a list of inexpensive tools for massive content-based text processing, including the Treebank corpus, PIC-Kimmo and Englex, Perl, SMART, and JUMAN. firstname.lastname@example.org, (908) 582-7550 Fax. [IRLIST, 6/1/93.]
I mentioned that the Canadian Armed Forces are purchasing their first computerized map-making system. George Williams worked for the supplier's parent company, Intergraph, in Huntsville before he joined Boeing's Advanced Computing Group there. He says that digital mapping is a big market for Intergraph, and a major field for software R&D and government contracts. Satellite and aerial sensing are generating increasing amounts of data for land management (e.g., lumber), petroleum and mineral prospecting, government and military applications, and scientific research. Some rudimentary AI work had been done, but there is room for much more. "This field has tremendous potential for expansion in the coming decades." [email@example.com, 6/2/93.]
(Remember the excitement over SRI's Prospector system? It found a molybdenum ore deposit worth millions, hidden under a tailings pile at a mine site. It's not clear to me why Prospector faded out, even though I was at SRI when it happened. USGS was providing the expertise, and seemed happy with Prospector as a tool for knowledge capture. Other companies wanted the inference engine but didn't have the expertise to build conflict- free Bayesian inference rules. Academics were more interested in newer inference algorithms, such as Judea Pearl's Bayesian trees or the Dempster-Shafer algorithm. Rene Reboh couldn't keep the project alive, and eventually joined Peter Hart at Syntelligence to work on insurance problems. Prospector was a good system, though, and could still be of use. Its main fault was that it had no geometric reasoning. All geologic inferences were made at a single point, or were repeated independently for multiple points. If a rule specified "near a fissure," someone had to enter whether that was true for the point. Map overlays were a reasonable solution, but the Lisp programmers had little GUI software for working with maps. They weren't able to pull out a second success to sustain corporate interest, and SRI's full-time researchers couldn't be supported on academic-style research grants. Someone at a university might have had better luck.)
Thang Cao Nguyen (firstname.lastname@example.org) has taken a summer job at Amoco Research Center (Chicago) to work on image analysis for nucleus and chromosome detection and classification. On the weekends he drives three hours back to UIUC (as cthang @uirvld.csl.uiuc.edu) to work on 3D vision for his degree. He's interested in software tools for vision problems.