Computing devices that use non-traditional methods (e.g., gestures) to interact with data are rapidly becoming more popular, for both casual and power users. Applications and user interfaces for such cases pose a fundamentally different set of expectations that traditional databases are not well-suited for: users have come to expect response-times that are predictable and nearly instantaneous. Interactions such as accelerated scroll pose drastically different workloads to underlying databases and may overwhelm them. Without keyboards, several database query operations are nontrivial: gestural inputs have no clear way to be mapped to traditional queries. The traditional query-result model is insufficient to deal with nuances of gestural interaction, such as rapidly changing inputs and sensitivity of interfaces. The traditional database query paradigm needs fundamental rethinking to support gestural interaction. The PI will work on the missing key components of the database stack for interactive and gestural workloads, collectively entitled "GestureDB". The proposed research will make data interaction accessible to an entirely new category of devices and users. Gestural interaction is rapidly gaining popularity where data access is key but there is limited keyboard access, e.g. users with disabilities and environments such as factory floors and laboratories. GestureDB's interaction-focused design is also expected to have a transformative impact in data-driven fields such as bioinformatics and "big data" analytics. Further, the interactive nature of GestureDB will make it an excellent platform for creatively engaging with both students and the general public.
With GestureDB, the PI will first build a new query model featuring the concept of a "query intent", and an expressive "gestural query language," allowing users to use gestures as the sole mode of data interaction. Second, he will work on methods for "intent interpretation," allowing the system to better recognize the user's query intent during the gesture. Third, the PI will investigate methods for "feedback generation," allowing the system to provide feedback during the gesture articulation. All components will be designed while keeping interactivity in mind, in order to maintain a low-latency loop and ensure a fluid user experience. While decades of research in databases have gone into making databases more performant, the focus has typically been on large-scale pipelines, and not end users. Research in human-computer interaction and visualization has recently been investigating data management concepts for user interfaces. GestureDB bridges this gap, and takes a new approach towards enabling interactive, gestural querying of data. Ideas presented in this proposal are not restricted to gestural interaction: several can be adapted to solve classically hard problems in traditional database querying as well. Enabling gestural interaction will transform the default, traditional modes of ad-hoc querying of databases. The proposed solutions can open up several new avenues of research in query interfaces, query intent interpretation and feedback generation, and will also inspire research at all levels of the underlying database stack. The mechanisms proposed will enable building of highly interactive applications, and have a significant impact in rethinking future database systems.
For further information see the project web site at: http://interact.osu.edu/gesturedb
