Construction and Usage of the Semantic Interaction Pipeline

Semantic interaction techniques in visual data analytics allow users to indirectly adjust model parameters by directly manipulating the visual output of the models. Many existing tools that support semantic interaction do so with a number of similar features, including using an underlying bidirectional pipeline, using a series of statistical models, and performing inverse computations to transform user interactions into model updates. We propose a visual analytics pipeline that captures these necessary features of semantic interactions. Our flexible, multi-model, bidirectional pipeline has modular functionality to enable rapid prototyping. This enables quick alterations to the type of data being visualized, models for transforming the data, semantic interaction methods, and visual encodings. To demonstrate how this pipeline can be used, we developed a series of applications that employ semantic interactions. We also discuss how the pipeline can be used or extended for future research on semantic interactions in visual analytics.