Pickup Usability Dominates - Human-Computer Interaction and Innovation in Handheld, Mobile and Wearable Technologies - page 55

get the best of both worlds by conducting experi-
ments indoors that attempt to mimic some of the
distractions from the real world, while maintain-
ing the controlled laboratory environment (e.g.
Lumsden, Kondratova, and Durling 2007).
In experiments users are typically required to
enter a set of phrases on devices to measure their
text entry speed. While there is no widespread
agreement on phrases that are used, (MacK-
enzie & Soukoreff, 2003) proposed a standard
set of short phrases that has been used by other
researchers and provides a valuable baseline for
comparisons. One problem with mobile phone
text entry is that it is often used for short casual
messages and testing with formal phrases from
a traditional text corpus is not appropriate (see
differences discussed above for those who have
experimented with formal/newsarticle English
and SMS). This is compounded by the original
multi-tap text entry approach and short length of
SMS messages leading to considerable use of,
often obscure, abbreviations that are not normally
found in a corpus. To address this, (How & Kan,
2005) developed a large set of phrases extracted
from SMS users' real text conversations. Although
somewhat skewed to local Singaporean phrases
and abbreviations (much of SMS speak is heavily
localised and even personalised within a group of
friends), the corpus is a valuable insight into the
language often used on mobile phones.
Finally, it should also be noted that entry
speeds of 33wpm for users when transcribing text
on desktop keyboards have been found to drop to
around 19wpm for composing new text (Karat,
Halverson, Horn, & Karat, 1999), so most results
from text entry experiments can be assumed to
be over-inflating speeds by around 40% as they
are typically based on transcription.

that are very close to desktop keyboards, through
slight variations, to radical novel interaction
designs. We have looked at different hardware
keyboard designs, different on-screen keyboard
layouts, handwriting-based approaches and more
novel approaches such as gestures. We have also
looked at ambiguous and unambiguous designs,
and the related approaches to disambiguation.
Much of the work reported has experimental
backing to show the potential benefits of each
approach. However, when comparing the wide
diversity of approaches in the literature to widely
available implementations on real devices, the
overriding message we see is that guessability,
the initial pick-up-and-use usability of hardware/
software, is paramount to success.
It is extremely hard to predict future trends
for mobile devices: while there is considerable
research showing the benefits and strengths
of different approaches, market forces and the
views of customers and their operators have a
major role in deciding which techniques become
widely adopted. Predicted gains in expert text
entry performance are of no use if people do not
understand how to use the text entry approach out
of the box. To this end, we see considerable scope
for entry methods that provide a smooth transition
from novice to expert performance: XT9™ is one
successful example of novice-to-expert support,
as users get faster they'll learn to be sloppier and
type faster, without necessarily being consciously
aware of why. Context-aware word completion
that learns about individuals is another area that
shows good potential: good for slow novice typists
as they start, but building context and personalis-
ing as they gain proficiency.
Finally, looking at current market directions
and the increasing desire to enter more text on
small devices, we see the 12-key keypad slowly
disappearing from phones to be replaced with
less number-centric entry methods. Despite its
sub-optimality and problems on small devices,
both market trends and some user tests point to
the QWERTY keyboard taking on this role, either
as a physical or an on-screen keyboard.

CONCLUSION AND
FUTURE TRENDS

This article has reviewed a large number of text
entry methods that range from standard methods