1. Optimizing the Keyboard PSY/ORF 322 Final Project Jessica Blankshain, James Ma, and Robert J. Moore

2. Purpose Develop a keyboard optimization algorithm. –Utilize Fitts’s Law and other aspects of the Card-Moran-Newell (CMN) model of the human processor. Demonstrate that neither the QWERTY nor the Dvorak keyboard layout is an optimal configuration of the keys on the three-row keyboard. –These layouts can be improved upon using quantitative methods.

3. History: QWERTY Old typewriters used typebars and had an “invisible” printing point, leading to frequent jams. There is speculation that they intentionally placed all the letters of the word “typewriter” in the top row for quick and easy demonstration.

4. History: Dvorak In the 1930’s, August Dvorak and William Dealey developed and patented a new layout, known as the Dvorak Simplified Keyboard (DSK). Designed to improve typing efficiency and user comfort. Reasoning for layout was strictly qualitative.

5. Methods: Scale Issues Why Not Examine Every Possible Permutation? At one keyboard per second, that’s 8.41*10 24 years. The sun will burn out after 5*10 9 of them.

6. Methods: Keyboard Population Rather than look at entire keyboards, examine the influence of specific keys in specific locations. There are only 30*30 = 900 key-position pairs. We can randomly populate a large number of keyboards, then see how certain keys in certain positions influence their performance.

7. Methods: Scoring Keyboards First, need a figure of merit for keyboard performance: –Compile a large text file and calculate how long it would take an average human to type that entire file using that keyboard. –These figures are derived from the CMN model and Fitts’s Law.

8. Methods: Applying CMN Used Fitts’s Law to determine how long it would take a typist to move a finger from one key to another. Determined inter-row and inter-finger transition times. From this data, could determine the time to transition from any finger in any position to any other. See paper for quantitative details.

9. Methods: Compiling a Text Sample ~500KB of raw text 80,720 Characters Sources: –Popular Novels –Famous Speeches –TV Transcripts –Rap Lyrics

10. Methods: Algorithmic Flow Simulate 90,000 keyboards. Expect to see each key- position combination 3,000 times. Enough data to determine the best possible key-position combination. “Lock-In” that key. Repeat, simulating only the remaining 29 keys. Continue until entire keyboard populated.

11. Results Outperformed QWERTY by 33.3%. Outperformed Dvorak by 6.2%.

12. Conclusions Our method was successful. While DSK is a quantifiable improvement over QWERTY, our model is more efficient than both. Extensions: Software’s robustness allows for keyboards customized to users or industries.

13. Naming MNT RSH = Mt. Rushmore