| Popis: |
A keystroke-level model of user performance was developed to predict the improvement in text generation rate with a word prediction system relative to letters-only typing. Two sets of model simulations were tested against the actual performance of able-bodied and spinal cord injured subjects. For Model 1A, user parameter values were determined independently of subjects' actual performance. The percent improvements predicted by Model 1A differed from the actual improvements by 11 percentage points for able-bodied subjects and 53 percentage points for spinal cord injured subjects. Model 1B employed user parameter values derived from subjects' data and yielded more accurate simulations, with an average error of 6 percentage points across all subjects. |
