Automatically generating personalized user interfaces with Supple
Download 5.78 Mb. View original pdf
|
| Fig. 16. Supple’s customization architecture. The user’s customization actions are recorded in a customization plan. The next time the interface is rendered (possibly in a differently sized window or on a different device) the plan is used to transform the functional specification into a customized specification which is then rendered using decision-theoretic optimization as before. As illustrated in Fig. 16, customizations are recorded as a customization plan and they are represented as modifications to the original functional specification rather than as changes to a particular concrete user interface. Specifically, changes to the presence or location of user interface functions are recorded as modifications to the structure of the functional specification while modification to the presentation of the interface (users choice of a widget or layout fora particular element) are recorded as interface constraints. The interface generation process is thus extended to include an additional pre-processing step, where the customization plan is applied to the functional specification. Only then, the customized functional specification is rendered as a concrete user interface. This approach allows customizations performed on one device to be reproduced on other devices, except in cases where equivalent widgets or layouts are not available on the novel device. Customization plans are editable by users, who may choose to undo any of the earlier customizations, and they can do so even out of order (unlike the typical stack-based implementations of undo functionality. If any of the later customizations depend on the earlier customization the user is attempting to undo, Supple will notify the user of the dependency thereby allowing her to either abandon the undo operation or undo all dependent customizations as well. The separation of customization plans from the actual interface representation, together with the ability to edit those plans, offers the potential for users to share and collaborate on their user interface modifications. K.Z. Gajos et al. / Artificial Intelligence 174 (2010) 910–950 929 Download 5.78 Mb. Share with your friends:
|