Abstract

Handwriting has been shown to be a useful input modality for math. However, math recognizers are imperfect, especially when recognizing complex expressions. Instead of improving the recognizer itself, we explore ways to best visualize the recognizer's output to help the user fix recognition mistakes more efficiently. To do this, we propose changes to the visual editing operations in MathDeck, a math-aware search engine and formula editor, as well as the addition of an n-best list of results for each symbol in the recognizer's output. We present two experiments to help us find good ways to help users fix errors in the recognizer, and to test whether these changes help novices input formulas more efficiently than they would if they did not have handwriting as an input modality. In the first experiment, users had the option to fix errors with an in-place drop-down menu of alternate symbols, a side symbol correction panel, or by typing the symbols themselves or dragging them from a symbol palette. In our experiment, most users preferred to fix the errors manually by typing the correct symbols or using the symbol palette. In the second experiment, participants entered formulas using handwriting and/or LaTeX. We found evidence that suggests that novices can input formulas faster when they have access to handwriting, but experts still do better when they can just type LaTeX.

Library of Congress Subject Headings

LaTeX (Computer file); Mathematical symbols (Typefaces)--Classification; Writing--Data processing; Optical pattern recognition; Human-computer interaction

Publication Date

5-2021

Document Type

Thesis

Student Type

Graduate

Degree Name

Computer Science (MS)

Department, Program, or Center

Computer Science (GCCIS)

Advisor

Richard Zanibbi

Advisor/Committee Member

Joe Geigel

Advisor/Committee Member

Reynold Bailey

Campus

RIT – Main Campus

Plan Codes

COMPSCI-MS

Share

COinS