|
The Personal Wizards project has created and developed the field of collaborative programming-by-demonstration, an approach to capturing know-how of computer-based procedures by combining recordings of experts performing a task.
Collaborative programming-by-demonstration (PBD} is a way of developing rich procedure models by recording how one or more experts interact with an application while performing a specific task, and combining these recordings into an executable, distributable model via an appropriate learning algorithm. Collaborative PBD therefore extends traditional PBD by learning from multiple demostration and combining the collective knowledge of multiple experts. This approach models how to perform the same general task under different environmental conditions.
Uses of the procedure models created via collaborative PBD range from authoring intelligent documentation, to creating automation tools that combine the knowledge of a group of experts (e.g., technical support professionals, system administrators, or developers), to building guided application walkthroughs.
The Personal Wizards project has developed new learning algorithms and user interface techniques in support of collaborative PBD. Innovation in the learning algorithm space include: SimIOHMM, the first algorithm to combine multiple recordings to produce a probabilistic model of the procedure; Augmentation-Based learning (ABL), the first incremental learning algorithm that allows seamless interweaving of demonstations and manual edits of the model; and Distributed ABL, that extends ABL by supporting a new model of learning by reconciling separately modified procedure models through a version-control-like environment. Innovation in the user interface space include guidance techniques to assist users performing a task, and visualization techniques for guided walktroughs and tutorials based on procedure models captured via Collaborative PBD.
Read More
Last updated 1 Feb 2007
|
