Abstract

Lack of timely access to information is a significant problem for the nearly 24 million blind or visually impaired (BVI) individuals in the U.S. and 285 million globally. Significant progress has been made in ensuring BVI individuals get the necessary accommodations they need. However, very little work has been done towards making critical materials accessible in real-time, especially for non-textual materials such as math expressions, graphical representations, and maps. Current methods for authoring, converting, and/or producing accessible versions of non-textual materials requires significant human, time, and financial resources. The process also requires people with rare and specialized knowledge such as braille transcribers and tactile graphic designers.

My research program is aimed at addressing these issues and is unified by two complementary strands of applied science, (1) multisensory information access, and (2) AI. In this talk, I will first introduce the notion of multisensory information access, with a focus on how it impacts the perception, cognition, and behavioral characteristics of humans. Then, using the findings as foundation, I will show how we can (and are) using AI to enable multisensory information access in educational and navigational settings to address the long-standing accessibility issues of BVI individuals. https://www.unarlabs.com/