SIGART recently mailed their Oct96 (!) issue of the SIGART Bulletin, to be followed soon by the Aug97 and Oct97 issues. The bulletin will (most likely) then cease publication, instead merging with ACM's glossy new _intelligence_ magazine (starting Jan98). Sample sections of "_intelligence_: New Visions of AI in Practice" will be distributed with the last two bulletins. The new magazine will cover "techniques and practice of developing intelligent systems" and "the practical, professional, and commercial side of AI." It is an ACM Track II publication meant to have wide appeal, and may eventually be a bi-monthly. Book reviewers are needed. (Reviews will no longer include conference proceedings or -- maybe -- textbooks, except on the SIGART website.) Christopher Welty, , is editor in chief. SIGART will continue to provide services and accept announcements via , . _intelligence_ will cost $39 for ACM members and $49 for non-ACM members, and will include SIGART membership. To subscribe, contact . [SIGART Bulletin, Oct96; and SIGART EIS.] (I presume that SIGART will still participate in the biennial AI Directory, jointly published with AAAI and others.)

Lee Giles has posted Journal Citation Reports' "impact factors" for various CS/AI journals in 1995. (1996 scores may be available later this year.) Impact factor measures frequency of citation for the average article, normalized by journal circulation. Cognitive Brain Research leads the pack in 1995, with an impact factor of 2.2 -- but only .88 in 1994. (If what you want is a lot of readers, a large-circulation magazine may be more useful than one with a high normalized citation rate -- but the latter would indicate a tightly knit community of scientists publishing useful articles. The best journals have both: Science magazine has a 1995 factor of 22; Nature, 27; Clinical Research, 58.)

Other leaders in AI -- excluding machine vision journals and EE/robotics -- are IEEE Trans. on Pattern Analysis (1.9); Neural Computation (1.7); IEEE Trans. on Neural Networks (1.6); Artificial Intelligence (1.6); Machine Learning (1.3); Neural Networks (1.3); and Knowledge Acquisition (1.1). Less-cited journals include Network-Comp. Neural; AI Magazine (.86); AI in Medicine; Expert System Applications; IEEE Trans. on Knowledge and Data Engineering; Int. J. of Approximate Reasoning; Pattern Recognition; Neurocomputing; Int. J. of Intelligent Systems; IEEE Expert (.60); Applied AI; Pattern Recognition Letters; AI EDAM; Engineering Applications of AI; J. of Automated Reasoning (.25); J. of Experience and Theory in AI; AI Engineering; Int. J. of Software Engineering Knowledge; Knowledge-based Systems (.17); AI Review; AI Applications; Applied Intelligence; Computers and AI (.045); and Neural Processing Letters (0.000).

In 1994, the top three AI journals were Neural Computation (3.1); IEEE Trans. on Pattern Analysis (2.0); and IEEE Trans. on Neural Networks (1.9) -- all very similar to 1995. Communications of the ACM was 1.8. Additional info is on ("What's Here: Citation index"). [Connectionists, 04Sep97.]

-- Ken