mobile text entry: methods and evaluation csci 4800 march 31, 2005
TRANSCRIPT
Mobile Text Entry The problem Approaches
T9 Fastap Other …
Evaluation How it was performed? Results Discussion and analysis
Key-Based Methods
Telephone Keypad
Multi-press method Two-Key T9 text entry method LetterWise
Fastap
Dependent Measures
Text entry speed is characters per second (CPS), which is calculated as:
CPS=Cn/Tc Words per min (WPM)=CPS*60/Wc A metric for measuring the overhead
involved in correcting errors is keystrokes per character (KSPC)
KSPC=Kn/Cn Ratio Kn/Kmin when KSPC>1
Evaluation Phase
Participant Details:
34 participants took part Filled questionnaires detailing their
experience with text messaging 4 categories were made i.e. beginner
(zero messages a week), novice (> 5 msg. a week), intermediate (5 to 15 msg. a week) and expert (> 15 msg. a week).
Evaluation 30 mins session over a 2 week period Participants were asked to correct any
errors and complete the task. Fill out a NASA TLX worksheet. (Six
categories: mental demand, physical demand, temporal demand, effort, performance and frustration level)
Each participant received different order of test sentences from each category
Interfaces
Nokia 82602 ( T9 method) Ericsson T10s3 (Multi-press with
timeout of 2 seconds) Fastap prototypes ( with only text
entry functionality)
Test Sentences
24 sentences were created in each of the 4 sentence categories.
The traditional sentences contained only lower case letters and dictionary words.
Non-dictionary sentences Abbreviated sentences Numeric sentences
Evaluation Phase
Initial Reaction Novice evaluation Expert evaluation
Evaluations was based on questionnaires Mean entry rates and Error rates
Initial Reaction Results 30 of the 34 participants completed the
task within the time limit. mean entry speed across all participants
interfaces and also per Interface. First Task: T9 interface was the fastest Second Task: Fastap interface was fastest Multi-press interface performed poorly in
both tasks
Summary of Mean entry rates and SD
Fastap 6.417 (1.705), 8.249 (2.255)
Multi-press 4.537 ( 1.307), 1.936 (0.96)
T9 9.688 (6.699) ,3.641 (2.582)
Error Rates
Fastap interface had the lowest mean keystrokes per character for the 1st task followed by multi-press (1.51) and T9 (2.34).
Fastap interface had the lowest mean keystrokes per character for the 2nd task followed by T9 (2.82) and multi-press (7.20).
Comparison of Usability The average text entry rate for the T9 interface, illustrate the
potentially fast entry speeds can be achieved with the interface, by expert users. The Usability of Interface was very poor. The second initial reaction task proved easier for the participants using T9, with most being able to complete the task, though with varying levels of efficiency.
The multi-press interface performed poorly in both tasks, especially the second initial reaction task.Numerical entry was the most difficult. For the first initial reaction task the multi-press was more usable than T9, as all participants were able to complete the task fairly efficiently.
The Fastap interface provided the best immediate usability during the initial reactions tasks, with all participants being able to complete both tasks.
Second initial reaction, Participants indicated that number buttons were hard to press, having tried to press them directly rather than chording the four upper-layer buttons surrounding them. This does not seem to have a significant effect on their performance.
Error Rates
The Fastap interface had the lowest mean keystrokes per character value (1.09), followed by T9 (1.79) and multi-press (2.07).
Multi-press Provided quite good immediate usability,
though it performed poorly when entering numbers.
Received low ratings on the subjective response measures with participants being less than satisfied with its performance
Performance gains could not be achieved with further training because of the basic inefficiencies of the method
T9
Displayed very poor immediate usability Training period improved T9’s
performance Performed extremely well with the
traditional, numeric sentences and with the non-dictionary sentences
Tended to be most error-prone when entering non-dictionary and abbreviated sentences which often required mode switches.