Browse Prior Art Database

Has Member Relationships in a Semantic Network Database

IP.com Disclosure Number: IPCOM000122016D
Original Publication Date: 1991-Oct-01
Included in the Prior Art Database: 2005-Apr-04
Document File: 3 page(s) / 164K

Publishing Venue

IBM

Related People

Begas, W: AUTHOR [+5]

Abstract

Disclosed is a method for implementing HasMember relationships in a semantic network database. The disclosed method takes advantage of the fact that many HasMember relationships are rarely followed. It is possible to identify these and implement them with cheaper but slower mechanisms. A considerable savings of storage space is obtained without causing applications to process the respective semantic network any differently than they ordinarily would.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 37% of the total text.

Has Member Relationships in a Semantic Network Database

      Disclosed is a method for implementing HasMember
relationships in a semantic network database.  The disclosed method
takes advantage of the fact that many HasMember relationships are
rarely followed. It is possible to identify these and implement them
with cheaper but slower mechanisms.  A considerable savings of
storage space is obtained without causing applications to process the
respective semantic network any differently than they ordinarily
would.

      Semantic networks consist of subjects and bidirectional
relationships between subjects that work like complementary
cross-references.  The basic facility for organizing semantic
networks is a grouping mechanism called the "collection."  One
subject, the "collector" for the collection, represents the
collection as a single concept. This subject has "HasMember"
relationships from itself to a collection of other subjects which are
members of the respective collection.

      Some collections play an official role in the organization of
semantic network databases.  These collections are called
"categories."  There is one special meta-category named "Categories"
which has as members all other categories.  That is, its members are
the collectors of all regular categories.  This is, in fact, how
collectors are known to be representing categories, and not ordinary
collections. Each of these categories has a name unique to itself in
the set of all the categories in Categories.

      Categories are different from collections in that by being
members of "Categories" they are part of entry paths into the
semantic network.  Nominally, all paths begin at "Categories."  Paths
then follow HasMember relationships from there to a category selected
by its unique name.  From there, paths may follow HasMember
relationships to the members of the selected category.  Since the
members of ordinary categories need not have unique names, the
specification of a member name may indicate that multiple HasMember
relationships need to be followed.

      Categories are also distinct from ordinary collections in that
the entry paths they provide are also used to assure that
relationship deletions never isolate a portion of the network from
access.  This is called "island prevention" and is done by insisting
that every subject in the semantic network have one consistent and
dependable access path.  The facility insists that every subject be
made a member of some category at the time of its creation.
Moreover, the HasMember relationship which expresses this
categorization cannot be deleted until that member is deleted.  Nor
can the category collector be deleted until all mem bers have been
deleted.  Thus, for the life of that subject, that subject can be
accessed through that category.  That category is called the
predominant category for the subject, since the subject may
subsequently become members of other categories, a...