Adding Semantic Information into Data Models by Learning Domain Expertise from User Interaction
This work addresses the problem of enhancing interactive visual analytic systems for users by allowing them to add semantic information through their expertise, though it appears incremental as it builds on existing semantic interaction techniques.
The paper tackles the challenge of incorporating user domain expertise into data models by inferring and appending new dimensions based on user interactions, enabling direct manipulation of data spatial organization to weight existing dimensions and create new ones, as demonstrated in a use case.
Interactive visual analytic systems enable users to discover insights from complex data. Users can express and test hypotheses via user interaction, leveraging their domain expertise and prior knowledge to guide and steer the analytic models in the system. For example, semantic interaction techniques enable systems to learn from the user's interactions and steer the underlying analytic models based on the user's analytical reasoning. However, an open challenge is how to not only steer models based on the dimensions or features of the data, but how to add dimensions or attributes to the data based on the domain expertise of the user. In this paper, we present a technique for inferring and appending dimensions onto the dataset based on the prior expertise of the user expressed via user interactions. Our technique enables users to directly manipulate a spatial organization of data, from which both the dimensions of the data are weighted, and also dimensions created to represent the prior knowledge the user brings to the system. We describe this technique and demonstrate its utility via a use case.