[unrev-II] Issues in Knowledge Representaion (was Re: XML limits)

From: Paul Fernhout (pdfernhout@kurtz-fernhout.com)
Date: Wed Apr 05 2000 - 05:25:30 PDT

  • Next message: Paul Fernhout: "[unrev-II] The DKR hardware I'd like to make..."

    Jack Park wrote:
    > So, why not let's enumerate the problems of knowledge representation we want
    > to solve, then start from there.

    Good suggestions.

    As a start, here are some issues inspired by William Kent's 1979 book
    "Data & Reality"'s index and examples of when they come up.

    * Entities (quantity, change, categories, existence)
    Ex. You're reading a mystery novel (building up knowledge structures in
    a DKR as you go along) and as the end it says "The butler did it". What
    happens to those structures? (Kent's example).

    * Separating symbols and things
    Ex. When I say "tomato" that isn't everything I know about Tomatoes or
    everything the DKR knows about them.... Somehow that symbol can access
    additional information. How is this relationship between symbol and
    related data maintained? (And of course, we have to acknowledge, you
    can't eat the symbol "tomato".)

    * Naming (uniqueness, scope, qualifiers)
    Ex. You say "tomato" and refer to a vegetable, and enter a reference
    into the DKR, I say "tomato" and refer to a set of nutrition information
    and try to enter that refernece in the same DKR. How do we keep these
    same named references referring to different things? And likewise, if I
    use the scientific name for Tomato (?) how do we have it point to the
    same thing as some other reference to "tomato".

    * Representing relationships (items, domain, role, issues if computed)
    Ex. You make a link to the Tomato nutritional information from a report
    in the DKR. How is this relationship maintained? Some of the vegetables
    you reference have pictures, how do you build this link (relationship)
    when you didn't plan for it? Some of the nutritional information about
    a tomato is calculated -- how do you distinguish this from other
    non-computed information?

    * Attributes to objects vs. relationships
    Ex. Tomato's have weight. If you are inventorying your tomato harvest in
    a DKR, should you have a tomato class which has a weight attribute, or
    should you model this had a tomato being in a relationship with a

    * Why data and its description can't really be separated
    Ex. You want to send someone you tomato nutritional information so they
    can store it in another DKR. But in order to use it, they need to know
    about what format you stored it in.

    * Handling types and categories and sets
    Ex. (Weak!) What does it mean that nutritional information is about a
    type of Tomato? Does that then mean nutritional information can't be a
    type of empirical observation? Can it belong to the set of things your
    friend Fred entered for you, while ate the same time being in the
    category of things Jim fact checked? How is all this represented?
    Are actegories just relationships in disquise?

    * Modeling -- both modeling reality and modeling data
    Ex. Simulating the movement of air over a wing vs. discussing the way
    you implemented this, and also realizing your simulation is not reality
    (the map is not the territory -- but it is its own territory).

    * Handling elementary data (example: images or integers)
    Ex. Should an image stored in the DKR be made of relationships of bytes
    (and those relationships of bits?) or should it just be an array of
    bytes on a hard disk. In either case, how do you know how to interpret
    those bytes back into an image?

    * Data integrity
    Ex. For a DKR on tropical medicine, how do you make sure the weights of
    tomato plant parts entered do not add up to more than the weight of the
    entire tomato plant?

    * Data processes (triggers, integrity constrains, search)
    Ex. When you enter a record for tomato RDA by weight into your DKR, how
    can the DKR be set up to automatically send email to Eric to let him
    know to look it over?

    * Handling multiple versions of things or patterns
    Over time, we are going to get better information about tomato
    nutritional values. How do we keep all these versions of information?
    Which one do I give you when you ask for nutritional information on the

    * Testing for sameness (equality)
    How do I know that this reference to tomato nutritional information
    points to the same "thing" that this reference to "red fruit"
    nutritional information points to?

    * Efficiency
    Ex. Same as handling elementary data, also how does how the system looks
    to a user differ from how it is implemented.

    These are just some of the types of issues. As you can see, "XML" has
    nothing to say about almost all of these issues.

    -Paul Fernhout
    Kurtz-Fernhout Software
    Developers of custom software and educational simulations
    Creators of the Garden with Insight(TM) garden simulator

    Get a NextCard Visa, in 30 seconds! Get rates
    as low as 2.9% Intro or 9.9% Fixed APR and no hidden fees.
    Apply NOW!

    Community email addresses:
      Post message: unrev-II@onelist.com
      Subscribe: unrev-II-subscribe@onelist.com
      Unsubscribe: unrev-II-unsubscribe@onelist.com
      List owner: unrev-II-owner@onelist.com

    Shortcut URL to this page:

    This archive was generated by hypermail 2b29 : Wed Apr 05 2000 - 05:31:17 PDT