[unrev-II] Yes, there are scaling constants to knowledge catagorization

From: N. C a r r o l l (ncarroll@hastingsresearch.com)
Date: Sun Apr 15 2001 - 18:45:28 PDT

  • Next message: N. C a r r o l l: "[unrev-II] "Other" categories are both realistic and beneficial"

    ----- Original Message -----
    From: Bernard Vatant <universimmedia@wanadoo.fr>
    To: <unrev-II@yahoogroups.com>

    [snip]

    > 2. Creating ad-hoc folders for new objects - or presumed new - reminds me
    of what I call the Open Directory syndrom. In ODP, there have been, are, and
    probably will be 7 URLs per category and 10 categories per editor, these
    figures having an amazing stability at any moment in the growth of the
    project. I noticed that a year ago when I was an editor, there were about 1
    500 000 URLs at the time, and now with 1 000 000 more, the figures are the
    same! I check it every month ... you can do it yourself at http://dmoz.org .
    These figures are all the most amazing that they are obtained against the
    guidelines for ODP editors, trying to refrain them from unleashed creation
    of too many categories, recommanding not to split a category in
    subcategories before it counts about fifteen or twenty links. What does that
    mean? Does the categorization of knowledge have a kind of dimensionality, a
    scaling constant or whatever? I would be curious to know if other types of
    organization of collaborative knowledge bases will lead to such constants?
    For example, in building semantic graphs, do people converge towards a mean
    incidence - that is the number of edges incident to a node - around 7 maybe?

    Yes, there are scaling constants to knowledge catagorization. The best known
    is probably the Resnikoff-Dolby 30:1 rule, which finds that historically
    users of indexing systems prefer, on the average, 29.6 items per category,
    with a range of 10 to 160 at the extremes, depending on dozens of factors
    such as user expertise, the usability of the organizing system, monitor size
    in the case of computers, degree of variety within a category, visual
    presentation of the information (including all the elements of typesetting,
    such as size, weight, spacing, color, offset, etc.), and -- once beyond
    computers -- all the visual, spatial, and kinesic cues we use to order our
    worlds. (The R-D rule applies back to Babylonian times, I believe.)

    The commonly-seen hexagonal layout in topic maps, while it may have a visual
    appeal a la Jungian mandala, or feel "right" based on patterns in the
    physical world*, seems to me a fairly impoverished structure. Even limited
    to a screen and a mouse, the human is capable of far more discrimination
    than 6 or 7 nodes, especially when you consider that, when viewing from
    above, humans are quite good judges of vectors, down to less than 5 degrees
    of arc. It just takes better visual design.

    * Mayan markets of 1,000 years ago were layed out, by the natural forces of
    transportation economics, in a large hub market ringed by 6 smaller
    marketplaces. Nature holds more.

    Nicholas

    --
    ________________________________
    Nicholas Carroll
    ncarroll@hastingsresearch.com
    Travel: ncarroll@iname.com
    http://www.hastingsresearch.com
    ________________________________
    

    ------------------------ Yahoo! Groups Sponsor ---------------------~-~> Do you have 128-bit SSL encryption server security? Get VeriSign's FREE Guide, "Securing Your Web Site for Business." Get it now! http://us.click.yahoo.com/2cW4jC/c.WCAA/bT0EAA/IaAVlB/TM ---------------------------------------------------------------------_->

    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: http://www.onelist.com/community/unrev-II

    Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/



    This archive was generated by hypermail 2b29 : Sun Apr 15 2001 - 19:09:00 PDT