1. TAUCHI – Tampere Unit for Computer-Human Interaction 1 Alternate Keyboards for Text Entry – And How to Evaluate Them I. Scott MacKenzie

2. TAUCHI – Tampere Unit for Computer-Human Interaction 2 Plan Background Keyboards Evaluation Case study

3. TAUCHI – Tampere Unit for Computer-Human Interaction 3 Background

4. TAUCHI – Tampere Unit for Computer-Human Interaction 4 Text Entry Research - Timeline Research Activity Year 19601970198019902000 HUMAN-COMPUTER INTERACTION GUIs Mouse input Direct manipulation MOBILE COMPUTING Pen-based input Handwriting recognition Email, SMS messaging Two-way pagers, mobile phones HUMAN FACTORS Office automation Word processing Document management Lots Little

5. TAUCHI – Tampere Unit for Computer-Human Interaction 5 Keyboards

6. TAUCHI – Tampere Unit for Computer-Human Interaction 6 Virtual vs Physical Keyboards Virtual keyboards –Aka “soft keyboards” or “on-screen keyboards” –Similar to clicking buttons in a GUI –Typically used with a stylus (but also with eye/head trackers and other input mechanisms) Physical keyboards –Desktop qwerty, miniature qwerty, mobile phone keypad, 5-button pager, 3-key date stamp, 1-key input, etc. Design Issues –Key layout, key size, key shape, number of keys, activation force, disambiguation, language modeling, word prediction, etc.

7. TAUCHI – Tampere Unit for Computer-Human Interaction 7 Keyboard Layouts – A Quick Tour Qwerty

8. TAUCHI – Tampere Unit for Computer-Human Interaction 8 Dvorak

9. TAUCHI – Tampere Unit for Computer-Human Interaction 9 Fitaly

10. TAUCHI – Tampere Unit for Computer-Human Interaction 10 Opti

11. TAUCHI – Tampere Unit for Computer-Human Interaction 11 Qwerty variations

12. TAUCHI – Tampere Unit for Computer-Human Interaction 12 ABC

13. TAUCHI – Tampere Unit for Computer-Human Interaction 13 Half Qwerty

14. TAUCHI – Tampere Unit for Computer-Human Interaction 14 Phone

15. TAUCHI – Tampere Unit for Computer-Human Interaction 15 Five-key pager

16. TAUCHI – Tampere Unit for Computer-Human Interaction 16 Three-key date stamp  Select _abcdefghijklmnopqrstuvwxyz Hello ther e

17. TAUCHI – Tampere Unit for Computer-Human Interaction 17 One-key input

18. TAUCHI – Tampere Unit for Computer-Human Interaction 18 Number-of-keys Continuum Number of Keys MoreLess Ambiguity Continuum

19. TAUCHI – Tampere Unit for Computer-Human Interaction 19 Ambiguity Ambiguity occurs if there are fewer keys than symbols in the language Disambiguation is needed to select the intended letter from the possibilities Phone keypad is a typical example ? RUNNER Or, is it SUMMER, is it STONES ? Demo: java Decode d1-wordfreq-phoneks.txt 10

20. TAUCHI – Tampere Unit for Computer-Human Interaction 20 Word Prediction Basic problem… Given some amount of preceding text, predict subsequent text Design issues Dynamic vs. static language model Word-level or phrase-level prediction Size of candidate word list Candidate word selection Improving performance Demo: java WordPredict d1-wordfreq.txt 10

21. TAUCHI – Tampere Unit for Computer-Human Interaction 21 Word Prediction Example Consider the word “vegetable” –How many and what keystrokes are required? 4 keystrokes 6 keystrokes

22. TAUCHI – Tampere Unit for Computer-Human Interaction 22 Disambiguation + Word Prediction Demo: java PhoneKeypad d1-wordfreq-phoneks.txt

23. TAUCHI – Tampere Unit for Computer-Human Interaction 23 Evaluation

24. TAUCHI – Tampere Unit for Computer-Human Interaction 24 Quick Example Opti Qwerty Is Opti as fast as Qwerty?

25. TAUCHI – Tampere Unit for Computer-Human Interaction 25 Opti vs. Qwerty Opti is faster, but only after about 4 hours of practice

26. TAUCHI – Tampere Unit for Computer-Human Interaction 26 Evaluation Research questions –Typically, something like… Is design A as fast/accurate as design B? Research questions come together as… –Independent variables, and –Dependent variables

27. TAUCHI – Tampere Unit for Computer-Human Interaction 27 Independent Variables These are the factors and levels in your experiment Examples

28. TAUCHI – Tampere Unit for Computer-Human Interaction 28 Opti vs. Qwerty Two Independent variables –Keyboard layout with 2 levels: Opti, Qwerty –Session with 20 levels: 1, 2, 3, … 20 Referred to as a 2 x 20 factorial design The 40 test conditions were given to all participants, thus we have a 2 x 20 within-subject design (i.e., each subject received all 40 test conditions) Note: within-subject design = repeated measures design (cf. between-subjects design)

29. TAUCHI – Tampere Unit for Computer-Human Interaction 29 Dependent Variables These are the behaviours you measure Examples

30. TAUCHI – Tampere Unit for Computer-Human Interaction 30 Speed as a Dependent Variable Relatively straight forward to measure Example... 1 2 3 4 1234567890123456789012345678901234567890123 the quick brown fox jumps over the lazy dog t = 60 seconds = 1 minute Number of characters = 43 Number of words = 43 / 5 = 8.6 Speed = 8.6 / 1 = 8.6 wpm Note: Definition of a word: “five characters, including spaces, punctuation, etc”

31. TAUCHI – Tampere Unit for Computer-Human Interaction 31 Accuracy as a Dependent Variable A bit trickier to measure Example... the quick brown fox the quixck brwn fox Presented text Transcribed text 1.How many errors? 2.What are the errors? 3.What is the error rate (%)? 2 (gee, that was easy) An “x” was inserted An “o” was omitted Easy? Try this one (next slide) ER = 2 / 19 = 0.105 = 10.5%

32. TAUCHI – Tampere Unit for Computer-Human Interaction 32 Accuracy is a Bit Tricky quickly qucehkly Presented text Transcribed text 1.How many errors? 2.What are the errors? 3.What is the error rate? 3 (that was a bit tricky) ER = 3 / 8.25 = 0.364 = 36.4% quic--kly qu-cehkly quic-kly qucehkly qui-ckly qucehkly qu-ickly qucehkly Hmm, let’s see

33. TAUCHI – Tampere Unit for Computer-Human Interaction 33 Accuracy is a Bit Tricky (2) quickly qx  ucehkly qucehkly Presented text Transcribed text Input stream Error rate = 36.4% KSPC = 10 / 8 = 1.2 Note: KSPC = keystrokes per character

34. TAUCHI – Tampere Unit for Computer-Human Interaction 34 Keystrokes Per Character (KSPC) KSPC is useful both as a characteristic of text entry methods and as dependent variable in evaluations of text entry methods KSPC as a characteristic –The average number of keystrokes to produce each character of text for a given language and entry method; e.g., KSPC  1.00 for the Qwerty keyboard KSPC  2.02 for multitap on a mobile phone KSPC as a dependent variable –A behavioural measure of the keystroke activity in entering text; e.g., (see previous slide)

35. TAUCHI – Tampere Unit for Computer-Human Interaction 35 KSPC Characteristics KSPC > 1 KSPC < 1

36. TAUCHI – Tampere Unit for Computer-Human Interaction 36

37. TAUCHI – Tampere Unit for Computer-Human Interaction 37 Case Study

38. TAUCHI – Tampere Unit for Computer-Human Interaction 38 Three-key Text Entry The basic idea…  Select _abcdefghijklmnopqrstuvwxyz Hello ther e Edit buffer “Virtual” keyboard Physical keys

39. TAUCHI – Tampere Unit for Computer-Human Interaction 39 Design Issues Cursor modes –“Persistent” = stays put after each entry –“Snap to home” = snap to SPACE after each entry SPACE position –Left, middle, right, etc. Letter order –Alphabetical –By probability of letters, digrams, etc. –Fluctuating (next slide)

40. TAUCHI – Tampere Unit for Computer-Human Interaction 40 FOCL FOCL = fluctuating optimal character layout Letters rearranged after each entry “Highly probable next characters” positioned closest to cursor Advantage: fewer keystrokes Disadvantage: increased cognitive load

41. TAUCHI – Tampere Unit for Computer-Human Interaction 41 FOCL “Level 1” and “Level 2”

42. TAUCHI – Tampere Unit for Computer-Human Interaction 42 Three-Key Method Comparisons The next step…

43. TAUCHI – Tampere Unit for Computer-Human Interaction 43 User Testing Methods #2 and #6 chosen for user testing

44. TAUCHI – Tampere Unit for Computer-Human Interaction 44 Participants and Apparatus Participants –10 paid volunteers (8 male, 2 female) –Ages 20 to 49 (mean = 30.1, sd = 8.5) –3+ hours of computer usage per day Apparatus (standard PC) –Input Keyboard keys (configurable) Most used “z” = left arrow, “x” = right arrow, “Enter” = Select –Output CRT display (next slide)

45. TAUCHI – Tampere Unit for Computer-Human Interaction 45 Display - Method #2 Virtual keyboard Presented text Transcribed text

46. TAUCHI – Tampere Unit for Computer-Human Interaction 46 Display - Method #6

47. TAUCHI – Tampere Unit for Computer-Human Interaction 47 Typamatic Cursor Keys Auto-repeat, or typamatic, cursor key behaviour possible Used at discretion of participant Spec’s –Delay = 176 ms –Repeat rate = 31.2 char/second

48. TAUCHI – Tampere Unit for Computer-Human Interaction 48 Procedure and Design One-hour session per participant About 25 minutes of text entry for each condition Order of conditions counterbalanced Instructions: Enter text phrases as quickly and accurately as possible. Ignore mistakes. Continue in the event of a error. A few practice phrases, then data collection Post-test questionnaire

49. TAUCHI – Tampere Unit for Computer-Human Interaction 49 Speed 9-10 wpm Difference not statistically significant

50. TAUCHI – Tampere Unit for Computer-Human Interaction 50 Comparisons

51. TAUCHI – Tampere Unit for Computer-Human Interaction 51 Accuracy Difference not statistically significant

52. TAUCHI – Tampere Unit for Computer-Human Interaction 52 KSPC Characteristic (given earlier)

53. TAUCHI – Tampere Unit for Computer-Human Interaction 53 KSPC Observed 1 1 includes typamatic keystrokes

54. TAUCHI – Tampere Unit for Computer-Human Interaction 54 KSPC Observed 2 2 excludes typamatic keystrokes

55. TAUCHI – Tampere Unit for Computer-Human Interaction 55 Typamatic Events More opportunity with Method #2, due to greater cursor distances

56. TAUCHI – Tampere Unit for Computer-Human Interaction 56 Keystroke Categories Method #2 Method #6

57. TAUCHI – Tampere Unit for Computer-Human Interaction 57 Participant Questionnaire

58. TAUCHI – Tampere Unit for Computer-Human Interaction 58 Conclusions Entry rates about 9-10 wpm in first 30 minutes Linguistic optimization (FOCL) offers little or no speed improvement, and increases user frustration

59. TAUCHI – Tampere Unit for Computer-Human Interaction 59 Future Work Improved keying techniques for Method #2 Optimize typamatic keying spec’s (delay interval and repeat rate) Other techniques for cursor control (e.g., wheel) 2-key, 1-key text entry techniques Improved tools for error analyses (next slide)

60. TAUCHI – Tampere Unit for Computer-Human Interaction 60

61. TAUCHI – Tampere Unit for Computer-Human Interaction 61 Thank You Referernces 1.MacKenzie, I. S. KSPC (keystrokes per character) as a characteristic of text entry techniques, Proceedings of the Fourth International Symposium on Human Computer Interaction with Mobile Devices. Heidelberg: Springer Verlag, 2002, 195-210. 2.MacKenzie, I. S. Mobile text entry using three keys, Proceedings of NordiCHI 2002. New York: ACM, 2002, 27-34. 3.MacKenzie, I. S., Kober, H., Smith, D., Jones, T., and Skepner, E. LetterWise: Prefix-based disambiguation for mobile text input, Proceedings of the ACM Symposium on User Interface Software and Technology - UIST 2001. New York: ACM, 2001, 111-120. 4.MacKenzie, I. S., and Soukoreff, R. W. Character-level error analyses for evaluating text entry methods, New York: ACM, 2002, 241-244. 5.MacKenzie, I. S., and Soukoreff, R. W. Text entry for mobile computing: Models and methods, theory and practice, Human-Computer Interaction (2002), [to appear]. 6.MacKenzie, I. S., and Zhang, S. X. The design and evaluation of a high-performance soft keyboard, Proceedings of the ACM Conference on Human Factors in Computing Systems - CHI '99. New York: ACM, 1999, 25-31. 7.MacKenzie, I. S., and Zhang, S. X. An empirical investigation of the novice experience with soft keyboards, Behaviour & Information Technology 20 (2001), 411-418. 8.MacKenzie, I. S., Zhang, S. X., and Soukoreff, R. W. Text entry using soft keyboards, Behaviour & Information Technology 18 (1999), 235-244. 9.MacKenzie, I. S., Zhang, X. I., and Soukoreff, R. W. Stylus tapping on a soft keyboard, Behaviour & Information Technology 18 (1998), 235-244.