A machine-learning apprentice for the completion of repetitive forms

摘要

The authors have developed a software environment in which workers can complete repetitive forms, and a machine-learning and prediction system that works within it. The nonintrusive assistant or apprentice provides viable default values for blank fields in a form, saving users up to 87 percent in keystroke effort and correctly predicting nearly 90 percent of the form's values. The system and prediction methods are active, yet not intrusive. Default predictions are always displayed, yet the user can override them easily with normal editing commands.