Xref: utzoo news.misc:5516 comp.text:7415 comp.ai:7822 news.software.b:5988 Path: utzoo!utstat!news-server.csri.toronto.edu!clyde.concordia.ca!uunet!wuarchive!zaphod.mps.ohio-state.edu!sol.ctr.columbia.edu!caen!math.lsa.umich.edu!math.lsa.umich.edu!emv From: emv@math.lsa.umich.edu (Edward Vielmetti) Newsgroups: news.misc,comp.text,comp.ai,news.software.b Subject: natural language recognition suitable for netnews? Message-ID: Date: 13 Oct 90 07:16:22 GMT References: <2459@van-bc.wimsey.bc.ca> <1617@chinacat.Unicom.COM> <45512@apple.Apple.COM> <1990Oct12.105710.29940@chinet.chi.il.us> Sender: usenet@math.lsa.umich.edu Followup-To: comp.ai Organization: University of Michigan Math Dept., Ann Arbor MI. Lines: 44 In-Reply-To: laird@chinet.chi.il.us's message of 12 Oct 90 10:57:10 GMT Cross-posted to a few unlikely groups because I don't know the sub-specialty that knows how to deal with this problem. Followups to comp.ai. In article <1990Oct12.105710.29940@chinet.chi.il.us> laird@chinet.chi.il.us (Laird J. Heal) writes: What really needs to be done is to redirect inappropriate articles into more appropriate newsgroups, as with the reposted articles comprising comp.archives. Not exactly. comp.archives does not view any article as "inappropriate". Rather it culls out particularly appropriate articles, dusts them off and cleans them up a bit, and places them into a new additional group. Inappropriate articles are silently ignored. The problem with doing this is threefold. First, I would not trust a program to automatically inspect an article for content and place it in a newsgroup. Second, I would not trust any old Joe on the net to put my article in any old newsgroup, and third, anyone or any group of people who would be trustworthy to catalog the articles would burn out within a very short period of time. I wouldn't want to have software automatically take articles and rebroadcast them around the net; that has too many opportunities for feedback loops. A better solution would be to retroactively cross-post the possibly interesting or appropriate articles into a new local newsgroup, which you could then read locally and have a tremendous competetive advantage over other people who are slogging through junk. I think it's possible to build an expert system (or whatever the current word for those things is these days -- knowlege-based system? natural language understander?) which finds things suitable for comp.archives, and by extension any other group with a similar narrow focus. At least I would hope so, after all I have a few thousand articles which a human system has found, something should be trainable to mimic what I have done. --Ed Edward Vielmetti, U of Michigan math dept moderator, comp.archives after 1 November 1990: