Abstract: This paper describes Snippets, a novel method for improving computerized data entry from paper forms. Using computer vision techniques, Snippets segments an image of the form into small snippets that each contain the content for a single form field. Data entry is performed by looking at the snippets on the screen and typing values directly on the same screen. We evaluated Snippets through a controlled user study in Seattle, Washington, USA, comparing the performance of Snippets on desktop and mobile platforms to the baseline method of reading the form and manually entering the data. Our results show that Snippets improved the speed of data entry by an average of 28.3% on the desktop platform and 10.8% on the mobile platform without any detectable loss of accuracy. In addition, findings from a preliminary field study with five participants in Bangalore, India support these empirical results. We conclude that Snippets is an efficient and practical method that could be widely used to aid data entry from paper forms.