Arnon Amir, Michael Lindenbaum
IEEE Transactions on Pattern Analysis and Machine Intelligence
How to index or retrieve multimedia objects is by no means obvious, because the computer can retrieve right multimedia material only if it reasons about its contents. We show that it is possible to write formal specifications of this reasoning process using set theory and mereology. We discuss the theoretical consequences of trying to use mereology and set theory for multimedia indexing and retrieval. We re-examine the roles of mereology and set theory in knowledge representation. We conclude that both commonsense set theories and mereologies should play the role of constraining databases of arbitrary multimedia objects, e.g. video clips. But although both should be viewed as database constraints, we argue that part-of hierarchies should be used to encode relatively permanent background knowledge, elsewhere names the referential level, while member-of hierarchies should describe arbitrary multimedia records. We also propose a language and a set of axioms, SetNM, for natural mereologies with sets. A multimedia indexing system can then be viewed as a particular SetNM theory. © 1995 J.C. Baltzer AG, Science Publishers.
Arnon Amir, Michael Lindenbaum
IEEE Transactions on Pattern Analysis and Machine Intelligence
Bing Zhang, Mikio Takeuchi, et al.
NAACL 2025
Barry K. Rosen
SWAT 1972
Arnold L. Rosenberg
Journal of the ACM