Chris Dent
L505 Essay 12
2001-04-09
Privileging the Hegemony of Hierarchy
The field of information science is filled with notions of hierarchy. Most representation schemes have a notion of a top, which contains only a few general items but items that have significant capacity to contain other items. Below those items are more items with less capacity trailing off in a recursive path of this is and this is not. There’s so much of this that it can probably be assumed that we, humans, are just built that way. We like structures that impose hierarchy; we want things in their place, preferably their one place.
Pity. Imposing and preserving hierarchy is such a conservative gesture. While it helps to prevent information overload it constrains perception, thus constraining new ideas. The point of excellent knowledge representation, after all the issues of improving accessibility and making sure it is fair to everyone, is making people more able to know more.
The discussion of thesauri and ontologies presented in the references below suggests a way to move imposed hierarchy out of the way until it is actually needed. The general notion of creating a resource tends to begin with the creation of broad categories of separation (e.g. physics, chemistry and astronomy) followed by the narrowing of those categories (thermodynamics, kinetics, optics)[1]. The process used in creating an ontology offers an alternative method. Instead of first defining categories, define relationships between things (such as is-a, has-a, part-of, like-a) then gather the things together and create the relationships.
Databases can provide the tools necessary to keep track of all the objects. Each object can be evaluated as it is entered into the database, independent of the hierarchy. Multiple relationships can be created regardless of one thing containing another. Inside the database this is a tangled web but proper data selection can expose only the relationships that are important to the user.
If at some point a notion of hierarchy is required it can be selected dynamically from the database: choose which type of relationship represents hierarchy, perhaps “is-a”. The top of the hierarchy becomes those objects that have no is-a relationship. Those objects to which no objects have an is-a relationship make up the bottom of the hierarchy. In between are objects connected in branching relationship chains. If a true directed acyclic graph[2] is required the computer can follow the relationships between the objects to rationalize the hierarchy. One thing that would need to be decided is how to contain something that upon entry gave itself an is-a relationship in such a way as to doubly nest objects. Does the lowest level object become a member of the object at the top of the nesting or the object directly above it? Such decisions would be dependent on the nature of the data.
Creating the hierarchy, though, may not be necessary. If some method of entry into the system is provided, that’s all that is really needed to then be able to navigate the relationships between the objects. The entryway does not have to be through some artificial notion of “the top”, it can simply be someplace relevant to the searcher.
REFERENCES
Batty, D. (1998). WWW : Wealth, Weariness or Waste: Controlled vocabulary and thesauri in support of online information access. D-Lib Magazine http://www.dlib.org/dlib/november98/11batty.html
Craven, T. (n.d.) Introductory Tutorial on Thesaurus Construction http://instruct.uwo.ca/gplis/677/thesaur/main00.htm
Palmer, M. ed. (1999). Excerpta from: "Multilingual Information Management
Current Levels and Future Abilities". http://www.formalontology.it/multilingual_information_managem.htm.
Qin, J., & Paling, S. (2001). Converting a controlled vocabulary into an ontology: the case of GEM. http://www.shef.ac.uk/~is/publications/infres/6-2/paper94.html.