Multimedia structuring using trees

Traditionally work on multimedia structuring has been centered on the creation of indices and their use for searching. Although searching is important there are many cases where the user just wants to browse through the data to find something interesting without having any particular search goal. Multimedia data exhibits hierarchical structure that can be exploited for more natural user interaction with the content. In order to handle the large amounts of multimedia data more structure than what is currently available is required. In this paper, we have focused on structuring multimedia data using trees to describe both temporal and categorical relations. The pervasive use of trees to express hierarchies facilitates browsing, profiling, and authoring. Our main target application is the implementation of a personalized TV−guide. The constraints imposed by this application caused the development of a new simple compact graphical user interface for tree browsing. Introduction There is a huge increase in the amount of multimedia data on the Web and elsewhere. This increase will continue as the television, video, and music industries gradually become digital. More advanced tools than what those which are currently available are required to handle this data efficiently. Structuring the information for browsing and retrieval is an important part of this process. File systems with directories were introduced in the early days of computing as a way of organizing the increasing amounts of computer files. In a similar manner trees are representations that can be used to hierarchically organize temporal multimedia data streams like video and audio. In this paper, we propose the use of trees as data representations that can be used for better browsing and analysis of temporal data streams. Multimedia data exhibits hierarchical structure at different time granularities. For example, a news broadcast is organized in sections, each section has different topics and each topic several scene changes or shots. As another example, music exhibits hierarchical structure at different time granularities. In our system the hierarchical structure of trees is used to express temporal and categorical relations and allows for more natural browsing, user−profiling and searching than the traditional tape−recorder user interface paradigm especially for large data sets. Traditionally, work on multimedia data has been centered on the creation of indices and their use for searching. Although searching is important there are many cases where the user just wants to browse through the data to find something interesting without having any particular search goal. In our system trees are not only used internally but are exposed and used extensively both by the multimedia author and by the end user. A prototype implementation of a system for authoring and viewing trees containing multimedia data is described. It is implemented within the Open Agent Architecture (OAA) (Martin et al, 1999), a distributed multiagent framework that enables rapid integration of component technologies. Issues related to the creation, presentation and maintenance of this representation are discussed. As an application of the system we describe a personalized TV guide. A novel graphical user interface for tree viewing on a TV screen using the discrete keys of a remote control has been developed.