Analyzing visual layout for a non-visual presentation-document interface

Abstract

Presentation documents play important roles in many fields, such as business and education. The principal purpose of presentation documents is to convey information visually, so recognizing the visual layout is essential for understanding those documents. However it is inherently difficult for the blind people to recognize a visual layout, because there are numerous types of charts in presentation documents. As the first step to solve such problems, this study focuses on diagrams in which objects or groups of objects are bound by arrows. Such diagrams usually show relationships among the objects. If such relationships could be recognized by screen readers, it would make them accessible. However, the presentation authoring applications do not have functions for embedding these relationships among objects. Therefore this paper proposes a visual analysis method for diagram structure in presentation documents to automatically create metadata. It generates metadata which describes the relationships of objects, and the source-destination relationships of arrows. Then a novel interface utilizing the metadata was prototyped to present the visual structure of presentation documents in a tree view. This allows blind users to understand presentation documents easily, because it represents the visual structure that current screen readers cannot expose. In addition, they are familiar with the tree view interface, so they can use it without training. Finally, an evaluation shows that our method for automatically creating the metadata can be applied to various types of diagrams in presentation documents. Copyright 2006 ACM.