KAV-DB : Towards a Framework for the Capture and Retrieval of Visualization Knowledge over the Web

Digital images have become a ubiquitous medium for information communication across a broad range of application domains. Along with this ubiquity has come an immense growth in the capabilities of the devices and tools used to produce and edit this imagery. New advances in algorithms and methods are made at a large pace, which makes staying on top of the learning curve difficult for general users and even experts. To accommodate this need, commercial devices and software packages typically provide a suite of presets and shortcuts for common tasks with intuitive descriptors, determined by extensive internal user studies and expert interaction but often without publishing the actual parameters and their settings. We describe an emerging framework which strives to externalize these practices into a centralized web-based community effort called KAV-DB (Knowledge-Assisted Visualization Data Bank), to allow coverage of algorithms and applications not currently driven by immediate commercial focus but of wide interest to the community of visualization researchers. The vision of KAV-DB is to provide a web service to capture, analyze, and retrieve parameter settings for visualization algorithms, given the data at hand. KAV-DB builds on a robust user study evaluation theory, called conjoint analysis, to formulate statistical models of method parameters extracted by ways of efficient user studies. We demonstrate the assessment and analysis stage our framework via two diverse example applications: relation-aware volume exploration and text annotation of color images. Digital Object Identifier 10.4230/DFU.xxx.yyy.p