Predictive Models of Form Filling

In this paper we investigate predictive models of form filling. A predictive model of form filling aims to reduce the amount of time a user spends filling out a form by predicting the values of fields on the form and using these predictions to make suggestions to the form filler. Existing predictive models ignore both the values of other fields on the form and the values previously entered by users other than the current form filler when predicting the value of a target field. We introduce a novel model of predictive form filling named the Collaborative and Contextually Frequently Used model that utilizes these sources of information. We demonstrate that our model outperforms the existing, standard models of predictive form filling using a real-world collection of forms. We also demonstrate that using the values used by other form fillers can significantly improve the performance of the existing models. Author