Abstract

We present a generic model for providing blind and severely vision-impaired readers with access to online information graphics. The model supports fully and semi-automatic transcription and allows the reader a choice of presentation mediums. We evaluate the model through a case study: online house floor plans. To do so, we conducted a formative user study with severely vision impaired users to determine what information they would like from an online floor plan and how to present the floor plan as a text-only description, tactile graphic, and on a touchscreen with audio feedback. We then built an automatic transcription tool using specialized graphics recognition algorithms. Finally, we measured the quality of system recognition as well as conducted a second user study to evaluate the usefulness of the accessible graphics produced by the tool for each of the three formats. The results generally support the design of the generic model and the usefulness of the tool we have produced. However, they also reveal the inability of current graphics recognition algorithms to handle unforeseen graphical conventions. This highlights the need for automatic transcription systems to return a level of confidence in the recognized components and to present this to the end-user so they can have an appropriate level of trust.