A user-driven annotation framework for scientific data

Annotations play an increasingly crucial role in scientific exploration and discovery, as the amount of data and the level of collaboration among scientists increases. There are many systems today focusing on annotation management, querying, and propagation. Although all such systems are implemented to take user input (i.e., the annotations themselves), very few systems are user-driven, taking into account user preferences on how annotations should be propagated and applied over data. In this thesis, we propose to treat annotations as first-class citizens for scientific data by introducing a user-driven, view-based annotation framework. Under this framework, we try to resolve two critical questions: Firstly, how do we support annotations that are scalable both from a system point of view and also from a user point of view? Secondly, how do we support annotation queries both from an annotator point of view and a user point of view, in an efficient and accurate way? To address these challenges, we propose the VIew-base annotation Propagation (ViP) framework to empower users to express their preferences over the time semantics of annotations and over the network semantics of annotations, and define three query types for annotations. To efficiently support such novel functionality, ViP utilizes database views and introduces new annotation caching techniques. The use of views also brings a more compact representation of annotations, making our system easier to scale. Through an extensive experimental study on a real system (with both synthetic and real data), we show that the ViP framework can seamlessly introduce user-driven annotation propagation semantics while at the same time significantly improving the performance (in terms of query execution time) over the current state of the art.