The paper discusses experiments involving a method for the automatic detection, prior to the integration of database schemas, of conflicts in the naming of data elements within these schemas. The method relies on the representation of semantic information (called quiddity) about the data elements present in the various schemas. The authors develop several inference procedures which, utilizing this information, determine whether two distinctly named elements in fact represent the same object, or if elements with the same name actually represent different objects. The experiments are concerned with (a) examining the accuracy and consistency with which quiddities of data elements might be declared by different database designers, and (b) evaluating the accuracy and errors of these automated procedures. The results indicate that the method has promise for use in detection of naming conflicts, and that certain inference procedures are superior to others in terms of their accuracy and error rates.
展开▼