Expanding Information Access through Data-Driven Design
MetadataShow full item record
Computer scientists have made progress on many problems in information access: curating large datasets, developing machine learning and computer vision, building extensive networks, and designing powerful interfaces and graphics. However, we sometimes fail to fully leverage these modern techniques, especially when building systems inclusive of people with disabilities (who total a billion worldwide , and nearly one in five in the U.S. ). For example, visual graphics and small text may exclude people with visual impairments, and text-based resources like search engines and text editors may not fully support people using unwritten sign languages. In this dissertation, I argue that if we are willing to break with traditional modes of information access, we can leverage modern computing capabilities from computer graphics, crowdsourcing, topic modeling, and participatory design to greatly improve and enrich access. This dissertation demonstrates this potential for expanded access through four systems I built as a Ph.D. student: 1) Smartfonts (Chapter 3), scripts that reimagine the alphabet’s appearance to improve legibility and enrich text displays, leveraging modern screens’ ability to render letterforms not easily hand-drawn, 2) Livefonts (Chapter 4), a subset of Smartfonts that use animation to further differentiate letterforms, leveraging modern screens’ animation capability to redefine limitations on letterform design, 3) Animated Si5s, the first animated character system prototype for American Sign Language (ASL) (Chapter 5), which uses animation to increase text resemblance to live signing and in turn improve understandability without training, and 4) ASL-Search, an ASL dictionary trained on crowdsourced data collected from volunteer ASL students through ASL-Flash, a novel crowdsourcing platform and educational tool (Chapter 6). These systems employ quantitative methods, using large-scale data collected through existing and novel crowdsourcing platforms to explore design spaces and solve data scarcity problems. They also use human-centered approaches to better understand and address usability problems.