Similarity and Analogy over Application Domains
Simone Diniz Junqueira Barbosa, Karin Breitman, Antonio Luz Furtado, Marco A. Casanova · 2007
Abstract. Databases, particularly when storing heterogeneous, sparse semi-structured data, tend to provide incomplete information and information which is difficult to categorize. This paper first considers how to classify entity instances as members of entity classes organized in a lattice-like generalization/specialization hierarchy. Then, it describes how the frame representation employed for instances and classes, as well as the closeness criterion involved in the classification method, favors the practical use of similarity and analogy, where similarity refers to instances within the same class, and analogy involves different classes. Finally, the paper argues that similarity and analogy facilitate querying semi-structured data. 1.