Talking Diagrams: Unlocking the Knowledge Contained in Diagram Data

The fact that a large proportion of online content is visual creates a barrier to accessing the internet for blind and partially sighted people. Automatically generating descriptions of visual content, in other words translating visual into verbal information, would do much to ameliorate this situation. In the general case, this is far beyond the state of the art, but is becoming feasible for sub-classes of visual information. This ambitious PhD project aims to develop a comprehensive methodology for generating verbal descriptions for a family of such subclasses, line diagrams, which account for a large proportion of non-photographic online visual content.