Newsgroups: comp.ai
Path: cantaloupe.srv.cs.cmu.edu!rochester!cornellcs!newsstand.cit.cornell.edu!newstand.syr.edu!news.maxwell.syr.edu!worldnet.att.net!howland.erols.net!ix.netcom.com!donli
From: donli@netcom.com (Li)
Subject: Search Engine Technique
Message-ID: <donliE5K3oz.Bxy@netcom.com>
Organization: Netcom On-Line Services
X-Newsreader: TIN [version 1.2 PL2]
Date: Thu, 13 Feb 1997 19:34:11 GMT
Lines: 26
Sender: donli@netcom19.netcom.com

Hi,

I notice DEC's AltaVista Search engine has a new feature, called
LiveTopic, which is to refine the first batch of search results.  But the
way LiveTopic works, it seems to me, if not unique from other search
engine's technique, nor is it that creative, for instance, it further
processes search according to the frequency of occurence of searched
word(s), it doesn't make much sense at all.  The quality of a page/article
cann't simply be determined by how often it uses relative word(s)
pertaining to a topic.  Or, these type of search engine also has other
built-in technique that evaluates the quality of a
site/page/article/others?  Anyone has better idea? 

Another technique, though extremely difficult, I think, would be extremely
useful, is a search engine that picks up three or four paragraphs of a
page/article/the_like at random, and evaluate the quality of them against
a chosen topic or against an inferred topic, and then determine the
overall quality of the whole piece, and make recommendation (report back). 
This may involve linquistics, predetermined creteria etc.  But, I do
believe this approach will be superbly superior to any existing ones. 

Just a thought.


Li, Chunshen (Don)
hmm@dc.net
