Re: [unrev-II] Jack's Use Case: Context-sensitive representation

From: Paul Fernhout (
Date: Fri Oct 27 2000 - 06:04:47 PDT

  • Next message: Paul Fernhout: "[unrev-II] Unique IDs (was Re: Jack's Use Case...)"

    Jack Park wrote:
    > The topic maps folks do this with a *public subject* which is being defined
    > as a registered URN. Registering URNs doesn't appear to be all that easy.
    > In a closed system, you just start with the number 0, grab a couple of
    > numbers for some reserved things, then start giving every concept a new,
    > unique number. Pointrel does this. I suppose you'll need to use long
    > integers to pull that off, but, what the heck. Problem is, you cannot share
    > with others; that requires grabbing numbers from some registry. Unique URLs
    > would do, I suppose. If you chose a unique URL, then, at that web page, you
    > can put some kind of verbiage, photo, movie, whatever, so that all viewers
    > grok the same thing.
    > I refer to this process as *semantic grounding*. In closed systems, it's
    > rather easy. In open systems, it's really hard.

    I've thought about this some in the context of the Pointrel system. My
    choice of 4 byte integers for IDs in those verions of the code is simply
    pragmatic for the systems I work on and disk space limits I have. There
    is no reasons the ID cannot be longer (or shorter). I've also thought of
    unique IDs for these nodes (and know a way to support this, since the
    Pointrel system uses an index file for efficient search and that could
    include a hash table or other lookup system for longer IDs that were not

    However, if you look at Kent's comments in "Data & Reality" on this
    relating to ROSE, his idea is that one would have private stores which
    only pass "symbols" across an interface to another system (and never
    relations or objects or executeables). I think this is a good way to go,
    but it requires a sophisticated import / export process. In effect, you
    are treating such stores as brains or minds, and they communicate by an
    agreed on language (XML, proprietary binary formats, whatever). (The
    issue is somewhat more complex when you discuss making backups of just
    part of the system).

    For example, this group is looking at import / export:

    The issue with transporting things with other than an intermediate
    language is that one still has to reconcile similar concepts which have
    been given different IDs independently or integration of the new data.
    For example, if you and I merge our email, we might find out that we
    have different unique IDs for the concept of "Doug Engelbart" which we
    constructed when we first heard of him. So -- a reconciliation process
    is still needed. Similarly, we may still have collisions -- either from
    a failure of the unique ID system (same random number, mistake for other
    ID component) or from importing an old version of the knowledge system.
    For example after you and I reconciled on a new ID for "Doug" (creating
    some new version of the system) I might import some old email I had
    archived and lost, and then I would need to reconcile concepts again. I
    guess what I am saying is it is not clear you can get away without
    reconciling concepts.

    Yes in theory we could have an agreed upon dictionary everyone shared,
    but I don't think this is 100% workable in practice. You can have some
    concepts most everyone mostly agrees on at the start, but you face both
    new concepts being created and the meaning of concepts drifting. What we
    all are creating is something much beyond a dictionary -- and even
    beyond an encyclopaedia.

    This makes a lot of sense when you think that the meaning of a concept
    is really in large part the relations it is involved in (and their
    relations). So, no two non-identical systems will apply the exact same
    meaning to any concept (even if differences in meaning are not usually
    distinguishable in practice).

    For just one example, when I refer to "Bill Gates" we may both think of
    the same person, but if you think of him mostly as being the world's
    richest entrepreneur building an enterprise from scratch and I think of
    him as someone given a million dollar trust fund at birth, having a
    prominent lawyer father, and a well connected mother (hobnobbing with
    IBM high ups), "Bill Gates" may have totally different meanings to each
    of us. For example, you may refer to things as a "like Bill Gates" to
    imply something about entrepreneurship and realizing the "American
    dream", I might refer to things as "like Bill Gates" to imply something
    about being born part of the elite and an inheritor of titled
    monopolies. Merging these concepts may be possible, but it may again
    change the meaning of relations to them.

    The notion of the interrelation of meaning and metaphor comes up here,
    but I'm not ready to pursue it in depth (Douglas Hofstadter is pursuing
    it though).
    I guess what I am saying is that if we communicate in text (as in
    email), we are using many metaphors. We aren't just pointing at
    extremely well defined concepts. Even when I point at a tree, I might
    mean "circle of life" (meaning the leaves and becoming fertilizer for
    the tree) and you might see "firewood". As I write this I realize that
    the whole notion of hyperlinking to a sense denies the notion of
    metaphor. Perhaps that is one reason I think the future may be more that
    of dynamnic search (guided by intent or context or hints) rather than
    explicit links. For example, see these people who are looking into
    search aspects:
    especially their Interspace Research Project:

    However I still find links appealing in the sense of building up
    knowledgebases. However, this issue of [metaphorically guided] search
    vs. [explicit] link is a very interesting one. And just because I want a
    system to use links internally to represent my changing knowledge base
    does not mean it is the best way to communicate. Let me present a
    challenge that makes the point: how do you hyperlink a poem for public
    display? Yes, creating links may be easy for you to do for yourself and
    your own interpretation, but how do you do it for others?

    Obviously if we are referring to concepts that are intended to point to
    binary things (like a web page version or an email that was sent) we may
    have less difficulty reconciling the meaning of a global unique ID for
    these. However, the emergent meaning around relations that link these
    emails to other deeper meanings
    may create a similar problem to the "like Bill Gates" problem above.

    I think the resolution to this is to have a sytem that can handle
    multiple meanings for concepts (merging and disentangling as needed,
    like in Kent's "the butler did it" example)-- but this is easier to say
    than implement.

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

    -------------------------- eGroups Sponsor -------------------------~-~>
    It's Easy. It's Fun. Best of All, it's Free!

    Community email addresses:
      Post message:
      List owner:

    Shortcut URL to this page:

    This archive was generated by hypermail 2b29 : Fri Oct 27 2000 - 06:13:34 PDT