hwb@beach.cis.ufl.edu (Howard Beck) (07/20/89)
What does hypertext have to do with AI? I see very little going on in hypertext systems since 1> The relationships among topics are hardwired by the hypertext document developers who manually create links and likewise 2> The user must manually search for information by selecting the appropriate links. In other words, the computer has very little to do with the organization and retrieval of information, let alone anything even semi-intelligent. So when I see advertisments for expert system shells including hypertext capabilities, my expectations are raised only to find that the hypertext portion has been used to create a glorified help section.
) (07/20/89)
hwb@beach.cis.ufl.edu (Howard Beck) writes: >What does hypertext have to do with AI? Last semester in an expert systems course I took, my professor (Larry Travis, at the Univ. of Wisconsin), predicted that expert systems, hypertext, and databases were all on a collision course and would merge into a common computer science technology. I can certainly see this, since the knowledge bases of expert systems and databases are fairly similar. As for hypertext systems and expert systems, I can certainly see a common area. One could (theoretically speaking) build a system which crossed the knowledge base with hypertext-like links. This would at least give a really fun way of browsing through a knowledge base. I could also imagine building a human-readable knowledge base (like a standard document, or paper) which had embedded rules within in, along with hypertext links. This way one could read through the human version of the document (the "formal specification," if you will), examine the rules being used to model the formal specs by activating hypertext buttons in the document (one could easily see which rules pertained to a particular portion of the document), and then follow links between related sections of the document to find related rules and descriptions, etc. This is just conjecture, but it would probably work. >I see very little going on in hypertext systems since 1> The relationships >among topics are hardwired by the hypertext document developers who >manually create links and likewise 2> The user must manually search >for information by selecting the appropriate links. In other words, >the computer has very little to do with the organization and >retrieval of information, let alone anything even semi-intelligent. This may currently be the manner in which hypertext systems are implemented. However, if we stretch our minds a bit, I am sure some more interesting schemes would arise. If you want some real AI in a hypertext system with present technology, I would think that the documents would need some embedded information built into them, such as the rules I described above or maybe just keywords. At the very least, then, one could implement some AI search heuristics that could semi-automatically browse through the hyper-document. >So when I see advertisments for expert system shells including >hypertext capabilities, my expectations are raised only to find that >the hypertext portion has been used to create a glorified help section. It seems to me this kind of hype is the case with a lot of commercial expert systems. Prof. Travis claimed that even though there were dozens of expert system books being published, he had only found two worth using in our course. Also, expert systems (like much of AI) have been oversold to the public, with the "features" advertised by the companies being far less valuable and easy than they would like you to believe. However, this seems to be the case with much commercial AI technology (witness the large commercial and military projects, which first claimed that speech recognition was easy, then claimed that expert systems would solve so many difficult problems, then said that stuff like "autonomous land vehicles" would be with us in such a short time, and now are going on about how neural nets are going to be so useful because they will be able to learn new problems; a helluva lot of hype, but not much has come of it). It does not surprise me in the least that hypertext has gotten mixed into this arena of big promises and little substance. But enough babbling.... .oO Chris Oo. -- Christopher Lishka ...!{rutgers|ucbvax|...}!uwvax!uwslh!lishka Wisconsin State Lab of Hygiene lishka%uwslh.uucp@cs.wisc.edu Data Processing Section (608)262-4485 lishka@uwslh.uucp "What a waste it is to lose one's mind -- or not to have a mind at all. How true that is." -- V.P. Dan Quayle, garbling the United Negro College Fund slogan in an address to the group (from Newsweek, May 22nd, 1989)
sme@ivax.doc.ic.ac.uk (Steve M Easterbrook) (07/24/89)
In article <20607@bikini.cis.ufl.EDU> hwb@beach.cis.ufl.edu () writes: > >What does hypertext have to do with AI? > The answer seems to be that the two fields are radically different approaches to the use of computers to assist human thought. The AI field attempts to model intelligence, and so produce a machine that can assist people by doing [some of] the thinking. On the other hand hypertext systems do not allow the machine to 'understand' the information it is storing, but rather assist by allowing one to clarify and organise one's thoughts. They attempt this by doing the clerical information-retrieval type jobs, but in a more natural (associative) way. Incidently, systems that simply offer browsing of someone else's hardwired associative net are not true hypertext. The ability to modify, by adding one's own notes and associations (links), would be required to call it a real hypertext system. There is plenty of scope for cross-fertilisation of ideas between the two fields, as both approaches are valuable. I can see a relationship between semantic nets and hypertext networks, and maybe if the machine was able to reason about the content and context of the hypertext nodes this would be a useful advance. Steve
ntm1169@dsacg1.UUCP (Mott Given) (07/24/89)
From article <435@uwslh.UUCP>, by lishka@uwslh.UUCP (Not an illusion!): > hwb@beach.cis.ufl.edu (Howard Beck) writes: > > It seems to me this kind of hype is the case with a lot of > commercial expert systems. Prof. Travis claimed that even though > there were dozens of expert system books being published, he had only > found two worth using in our course... I am very curious to about which two books Prof. Travis espouses for studying expert systems, and why. Could you tell me what their titles are? -- Mott Given @ Defense Logistics Agency Systems Automation Center, DSAC-TMP, Bldg. 27-1, P.O. Box 1605, Columbus, OH 43216-5002 INTERNET: mgiven@dsacg1.dla.mil UUCP: ntm1169@dsacg1.uucp Phone: 614-238-9431 AUTOVON: 850-9431 FAX: 614-238-3214 I speak for myself
smb@datran2.uunet (Steven M. Boker) (07/28/89)
I am involved in a project here at Data Transforms that is attempting to create a hypertext grouping system that uses the notions of fuzzy logic to create the inclusion/exclusion relationships for the hypertextual links. We are hoping to create a system that will allow the hypertext links to be formed at runtime based on the needs of the user. We are stucturing each concept with weightings to prerequisites, correlates and consequences to create a time ordered knowledge network that the user can query. We hope to end up with a system that will provide the user with ways to find the information he needs to understand a concept even when he doesn't know which information he is missing. I don't know when or if this work will be commercially available, but I have found it fascinating if somewhat complex. I would welcome comments or other peoples experiences with this kind of knowledge representation system. Steve Boker smb@datran2.uunet
jps@cat.cmu.edu (James Salsman) (07/28/89)
In article <406@datran2.uunet> smb@datran2.uunet (Steven M. Boker) writes: > I would welcome comments or other peoples experiences with this kind > of knowledge representation system. IMHO, one of the things that the field really needs to concentrate on is full-text *DISAMBIGUATED* storage. What I mean by disambiguation is this. When I ask for a search on the key "Cardinal" I want the system to ask me: "Would you like information on: (1) Birds (2) Baseball or (3) Religion?" In other words, each key in the index should have a pointer to the precise usage that the text it indexes uses it as. Extra credit for parsing that last sentence. :James -- :James P. Salsman (jps@CAT.CMU.EDU)