Context-Aware Form Filling for Web Applications

Context-aware user interfaces facilitate the user interaction by suggesting or prefilling data derived from the user's current context. This raises the problem of determining which context information can be used as input for which interaction element in the user interface. This task is especially challenging as the texts that describe the elements, e.g. their labels, often differ in the terminology used. In this paper, we present a novel mapping process for that purpose which (i) combines the advantages of string-based and semantic similarity measures to bridge the vocabulary gap between context and UI\ element, and which (ii) is able to automatically extend its vocabulary by observing the user's interactions. We show that these two features dramatically increase the quality of the resulting mapping. Unlike previous approaches, the proposed mapping process does not require any training or manually tagged data. Further, it does not only use the label to describe the context and UI elements, but additional texts like their tooltips.