Using graphical representation of user interfaces as visual references

[摘要] My thesis investigates using a graphical representation of user interfaces - screenshots - as a direct visual reference to support various kinds of applications. We have built several systems to demonstrate and validate this idea in domains like searching documentation, GUI automation and testing, and cross-device information migration. In particular, Sikuli Search enables users to search documentation using screenshots of GUI elements instead of keywords. Sikuli Script enables users to programmatically control GUIs without support from the underlying applications. Sikuli Test lets GUI developers and testers create test scripts without coding. Deep Shot introduces a framework and interaction techniques to migrate work states across heterogeneous devices in one action, taking a picture. We also discuss challenges inherent in screenshot-based interactions and propose potential solutions and directions of future research.

[发布机构] Massachusetts Institute of Technology