Supporting the selection of web content modality based on user interactions logs

Abstract

Internet offers a huge amount of services related to government, entertainment, and communication. However, to ensure that any citizen can have proper access to all services and information, it is necessary to think about accessibility in a broader way than it is found in the Web. Universal Design points to a way of thinking about the heterogeneity of capabilities instead of segregating according to individual disabilities. In the Web, applying this approach involves solving the challenge of selecting the appropriate content modality, at the appropriate time. Thus, this work present a study focusing on selecting the modality of a Web content based on the data resulting from a single pageview, minimizing efforts related to profile filling or explicit customization of Web systems. The study involved the analysis of 261 sessions, 31 coming from people that informed being using assistive technology. From the experiment performed it was possible to identify that the algorithm Random Committee provided the best results for the studied problem (precision 94.8%; recall of 94.3%). Results are potentially of interest for Human-Computer Interaction practitioners aiming at providing adequate content modality after minimum usage.