Automatic interpretation of proofreading sketches (Rodríguez, Sánchez & Lladós)

Summary:

This paper describes a sketch-based system for proofreading documents. The authors focus on creating a natural interface that does not force the user to use specific gestures, type to insert words, or click between gestures. They implement a set of gestures to represent some of the simpler actions based on proofreading conventions. They also allow the operations replace, delete, and insert to be specified using any symbol, where the symbol is drawn at the point in the text to be affected and the changes are written in the margin accompanied by the same symbol. Strokes are marked as connected if they are close enough together, and the page is divided into sections to help limit the context that needs to be considered. They give an extensive formal description of their symbol recognition processes. They included a commercial handwriting recognizer. They performed a user study with 9 proofreaders who filled out 33 documents and found that 90.1% of the annotations were correctly recognized, and they believe a method for discrimination between symbols and words would improve on this.

Discussion:

I like the idea of allowing the user to create her own symbols, especially applying to a local context of a page. I somewhat remember a paper we read in IUI dealing with annotations and highlighting, and I think it could be useful to combine the ideas and get a system that lets you define symbols to indicate level of importance or relation to a particular topic, and then retrieve all pieces of highlighted/annotated material marked with that symbol (in some classes, for example, I put stars in my notes next to things that I expect may be on an exam, or make other symbols for other meanings.)