Icons are prevalent from road signs to graphical menus, conveying complex concepts with simple graphical representations. However, designing such icons, particularly for a compound concept such as health insurance, requires sophisticated and creative design skills to effectively combine semantics, layouts/space and styles. In this study, we aim to automate the generation of icons given compound concepts. Informed by interviews with professional icon designers, we developed Iconate, a novel icon design tool that automatically generates icons based on textual queries and allows users to explore and customize the icons. We also introduce a computational pipeline that automatically finds typical icons for sub-concepts, and arranges them according to inferred conventions. To enable the pipeline, we also collected a new data set consisting of concepts and annotated icons. We evaluated the feasibility of the tool with designers and the validity of the pipeline through a crowd sourced experiment.

‍

The paper can be found here (pdf here). Our presentation video is on YouTube here, the project webpage is here, and the interface we created is here.

‍