Characterizing Multimedia Semantics
Today, the Internet hosts many image and video collections related to current news, sports, medicine, digital heritage, and other topics in the public domain. This vast collection is a huge source of knowledge and can be used in various ways, from searching for some music for entertainment to learning the characteristics of a speciﬁc type of tumor for research purposes. Discovering, correlating and characterizing the relevant media instances in a speciﬁc context is, however, a daunting task. Generally, a semantic description of the media contents, or what they signify, is more important to a user than its physical format and visual attributes. The challenge of imprecise association between the media attributes and their connotations is compounded by the diﬀerent contexts of media usage and the variety of media forms.