Table Structure Recognition and Its Evaluation

Tables are an important means for communicating information in written media, and understanding such tables is a challenging problem in document layout analysis. Our earlier paper addressed the problem of detecting tables in multiple media. The current work builds on our earlier work by recognizing the structure of the detected table. This forms part of a prototype we have implemented for a complete, end-to-end table understanding system that takes raw ACSII text as input, detects and parses any number of tables that it might contain, and generates a simple interactive man-machine dialog allowing a user to access the table data via spoken language interface.