Analysis of “MathBrush: A Case Study for Pen-based Interactive Mathematics”
Comments Made Elsewhere:
Summary:
Describe MathBrush, a pen-math tool for recognizing hand-drawn math expressions, using representation in MathML, recognition for specific users, among other features. Sits on top of a CAS (Computer Algebra System). Just there to recognize and manipulate, not calculate. This paper describes the rationale of MathBrush in light of education.
Had many unique design challenges since they were first to integrate with a back-end CAS and support multiple and repeated operations on the input. (1) Display of complex of expressions and outputs, with scrolling canvas to be like paper. (2) Use of context-sensitive menus to access commands. (3) Editing of subexpressions (i.e. edit only part of the equation) in a separate window. (4) Graphical plotting. (5) Other features not tested in their evaluation…
Did think-aloud, semi-structured interviews. Had to correct for people leaving extra ink on the equation. Had a two-tier recognition phase (character, then structural) that confused people when the latter would fail (assumed everything was ok after the first passed). Had trouble conveying what needed to be corrected to the user when an error occurred, resulting in the user making lots of unneeded corrections.
Noticed that users were changing their writing style to the system. Questioned whether their overhead was better than using pen or paper.
Discussion:
It’s funny that a participant suggested adding a on-screen keyboard to the application. People probably generally feel this is faster.
In this situation, I think it’s a good use of context menus as they visually free the interface and don’t require the user to learn a gesture for initiating commands. However, I am a fan of the Microsoft Office “ribbon” and I would not be opposed to having something like this that changed accordingly for the user.
They might have borrowed directly a CAS too much, in that they displayed the input and output instead of just doing correction. Mathematica and others work like MathBrush does, but they expressed issues with participants trying to edit the output instead of their gesture, and also people not knowing where to start when an error occurred. Perhaps “beautification” as the user went along would be better, and then making the equation red until it passed structural analysis (when it could turn green or something).
No real complaints on the people adjusting their writing style to the system (though they did have to slow down). There will always be some of this, but don’t exceed an extraneous amount.
Implementing “beautification” as the user inputs and having the equations red until they passed structural analysis are very good form of feedback to the user inputs. This will greatly enhance the usability. That was a good idea …