Simulating question-based visual scanning for non-visual readers
Date
2019
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
This dissertation describes the creation of a system for locating information in a text document that is relevant to a complex question. While the system can be used by anyone to efficiently identify text areas related to a question within a large amount of less relevant text, it was specifically designed for non-visual readers, notably people who are blind and low-vision. Visual readers often quickly scan through documents to locate relevant information, yet non-visual readers have few options for intelligently scanning documents for information relevant to complex questions. This can reduce efficiency in answering homework questions, in obtaining relevant information in work documents, in learning new information for enjoyment, and even in retrieving information from a previous pass through a text document. Thus the purpose of this dissertation is to develop a system that allows non-visual readers to gather information related to complex questions as quickly and efficiently as their visually-scanning counterparts. ☐ This dissertation is subdivided into 3 parts: 1. Determining how visual readers scan through documents when answering complex questions; 2. Developing and implementing a method that replicates the identification and location of relevant text within a document similar to what visual readers identified as determined by part 1; and 3. Developing a user interface to allow users to move through a document so that they garner all or most information garnered by visual scanners, including not only question-relevant information, but also topological information and information surrounding the relevant information.