Similarity-based retrieval of temporal documents

Abstract

In this paper, we describe a similarity-based retrieval framework that addresses the challenges associated with the temporal nature of multimedia documents. Multimedia documents consist of multiple media objects and a set of specifications (eg. temporal) that tie these objects together. Therefore, we describe similarity/dissimilarity measures that aim to capture document authors' intensions. We use a prioritized constraint-based framework to evaluate these measures. We also develop algorithms that efficiently compute these measures for special cases.