Improved taxonomical organization for searching in object-oriented databases

The ever-increasing amount of information which can be accessed on CD-i, CD-Rom, Bulletin Boards and telecommunication services makes searching for a particular piece of information problematic. Among other techniques, taxonomical organization is an effective method for tackling size and complexity, but for very large databases, deficiencies become apparent. A major problem is that either the depth of the tree or the branching factor becomes too big. Other problems are maintainability and artificial classifications. This paper discusses an approach wherein classes are merely used as aliases for a set of constraints on explicitly defined attributes. This enforces subclassifications to be formally related to each other and to be arrangeable in accordance with various mutually independent dimensions. At any node the user has the choice of which dimension to explore next. After the addition of new constraints, all subdimensions whose preconditions are rendered valid by this activity are presented and the process continues recursively until the user does not want to narrow down the search space along this dimension any further. The result is a much more structured search process, while the old-style taxonomy class labels, if defined, can be utilized as short-cuts.