Abstract:
We describe a technique that allows end-users to specify automated transformations of structured text by inferring an underlying model. Inference is achieved with a novel algorithm, Structured Prediction by Partial Match (SPPM), a generalisation of the well-known PPM approach to predictive text entry and compression. We created two simple applications, as examples of "first steps" end-user programming tasks that can be achieved using SPPM. In empirical evaluations, each of these applications proved to be substantially superior to equivalent facilities in leading commercial products.
PPIG 2008 - 20th Annual Workshop
Structured Text Modification Using Guided Inference