Loading ...
Sorry, an error occurred while loading the content.

1448Hierarchical relationships [Was visualization technologies]

Expand Messages
  • Leonard Will
    Jul 19, 2006
    • 0 Attachment
      In message <NDEFLGFFOLCGIGCFNPJKCEGLDBAA.bbater@...> on Tue,
      18 Jul 2006, Bob Bater <bbater@...> wrote
      >I can accept that some hierarchies demand or imply inherited
      >attributes. But as I suggested in my preceding posting, some
      >hierarchies don't (escalators). I would be very interested to hear from
      >anyone on this list who has encountered this issue and has managed to
      >make some sense of it. In the meantime, I shall leave the question open
      >and assume that some hierarchical relationships require inheritance,
      >and some don't. It would be useful to have a way of indicating, through
      >relationship naming conventions, whether a hierarchy implied or
      >discounted inheritance.

      The trouble with the things that are popularly called "taxonomies" is
      that they generally obscure or mix up the basis on which their
      hierarchies are constructed. I therefore prefer to talk about two
      better-defined structures: the thesaurus and the faceted classification
      scheme. "Taxonomies" often adopt some of the principles of these, though
      in an unsystematic way.

      1. Thesaurus
      In a thesaurus, hierarchies are built on the types you relationship you
      mentions previously: generic, partitive and instantial. These
      relationships apply only between concepts within the same fundamental
      facet or category, such as objects, people, activities, abstract
      concepts, places, and so on. When the relationship is generic or
      instantial, inheritance applies: the narrower concept inherits all the
      properties of the broader concept.

      The partitive, or part/whole relationship is restricted to a few special
      cases, such as parts of the body, disciplines, geographical areas and
      the structures of social organisations such as military units. It is
      dangerously misleading to define handrails or steps as parts of
      escalators, because there are many handrails and steps which are not
      parts of escalators. When dealing with objects I generally find it
      preferable to use a structure such as the following:

      escalators and escalator components
      - escalator components
      - escalators
      - - aluminium escalators
      - - wooden escalators

      The term "escalator components" can then be assigned to documents in
      addition to the general component term such as "handrails", "steps" or
      "motors". The British Standard for thesaurus construction,
      BS8723-2:2005, para. 7.3.3 d), advises that a "complex concept should
      usually be split if the term's focus represents a properly, part or
      component of the difference", so it is better not to use terms such as
      "escalator handrails", "escalator motors" and so on.

      A thesaurus built in this way is appropriate for post-coordinate
      searching, where concepts are combined in a search statement such as

      "handrails AND escalator components".

      2. Faceted classification
      Concepts in a classification scheme are typically "pre-coordinated", so
      that compound concepts are built at the time of indexing rather than at
      the time of searching.

      In that case concepts from different facets can be arranged one below
      the other in a kind of hierarchy, but in which the relationships are not
      the same as the thesaurus BT/NT relationships. It is best if node labels
      are included explicitly in a display to show where facets change, for

      mountain bicycles
      racing bicycles
      road racing
      cycle mechanics

      Here node labels containing facet names have been shown in angle
      brackets. These are not classes to which documents would be assigned,
      but just show how the structure has been constructed.

      In this structure, there are some hierarchical relationships, where (a)
      "bicycles" is a broader term of "mountain bicycles" and "racing
      bicycles", and (b) "road racing" is a narrower term of "racing". In each
      case the concepts in these relationships belong to the same facet, of
      "objects" in the first case and "activities" in the second.

      When a change of facet occurs, the relationship is not one of
      subordination but of coordination. The two occurrences of "manufacture",
      for example, are interpreted as labelling the compound concepts

      mountain bicycles : manufacture
      racing bicycles : manufacture

      the occurrence of "cyclists" here represents the compound concept

      racing bicycles : road racing : cyclists

      and the two occurrences of "wheels" represent the concepts

      mountain bicycles : wheels
      bicycles : wheels

      The concepts of "manufacture", "cyclists" and "wheels" could occur in
      many other contexts in the classification. They might also occur in
      their own right, uncombined, for general treatments of these topics. To
      identify these compound concepts we can either use a string of terms,
      linked by symbols such as colons as I have done above, or a symbolic
      notation which may be more concise and which may use distinct symbols to
      show which facets are being combined, as is done in the UDC [Universal
      Decimal Classification]. A symbolic notation is not essential for a
      classification scheme, and in schemes called "taxonomies" it is often
      dispensed with, the strings of terms being used instead.

      In records management, facets are often combined in the order
      <functions> : <activities> : <topics/transactions/tasks>. This sequence
      has been found appropriate for this application, but different
      sequences, or "citation orders", may be appropriate for other types of
      classification. For example, the Classification Research Group suggest
      that the following order will often be the best, particularly for
      technical subjects:

      thing - kind - part - property - material - process - operation- patient
      (i.e. system operated on) - product - by-product - agent - space - time

      As this is normally applied within a specific discipline, we can think
      of the "discipline", "subject area" or "function" facet as preceding
      this list.

      I'm sorry that this message has turned out rather long, but Bob did ask
      for a fuller discussion! I hope that this is of some help, and would be
      interested to know whether other people agree with my interpretation.

      Leonard Will
      Willpower Information (Partners: Dr Leonard D Will, Sheena E Will)
      Information Management Consultants Tel: +44 (0)20 8372 0092
      27 Calshot Way, Enfield, Middlesex EN2 7BQ, UK. Fax: +44 (0)870 051 7276
      L.Will@... Sheena.Will@...
      ---------------- <URL:http://www.willpowerinfo.co.uk/> -----------------
    • Show all 27 messages in this topic