Knowledge-based authoring tools (KBATs) for graphics in documents

Documents that contain both text and graphics are difficult to characterize for the purposes of retrieval and navigation. Digital Libraries of the future will be little advanced over current libraries if we cannot meet the challenge of dealing with the conceptual content of documents. There have been substantial advances over the last few decades in the parsing of text and graphics, but the methods are hardly ready for large-scale application. We submit that large strides can be made by building systems that allow the author to contribute to the generation of conceptual information during the authoring process. Thus, the information would be captured at authoring time rather than being laboriously reconstructed after the fact. The tools for doing this are called KnowledgeBased Authoring Tools (KBATs). We are building such a tool for creating gene diagrams and have elsewhere discussed the issues involved in designing KBATs for text (Futrelle & Fridman, 1995).