Savanta—search, analysis, visualisation and navigation of temporal annotations

Abstract

In this article, we present Savanta, an information gathering interface for temporal, semantic video annotations. In Savanta, we integrate various methods and paradigms for information gathering, including visualisation, filtering, data mining, navigation and search, in order to explore the possible advantages of doing so. We posit that a seamless integration of multiple access methods, combined with an improved interval visualisation scheme and dynamically generated metadata, will result in greater user satisfaction compared to conventional approaches for searching and querying video databases--despite the increased complexity that may result. We perform a formal usability evaluation comparing Savanta to systems based on traditional search/query paradigms, and conclude that Savanta outperforms them with regard to both power and usability, especially for complex and open-ended tasks.