T ext entry is a key component of many smartphone applications. The recent release of smart watches has met considerable interest, but without text entry, interaction is frustratingly limited-users can see posts, short messages, and emails but can't reply using the watch. As part of our ongoing work, we outline a text entry approach for smart watches, describe our initial prototype, and discuss the outcomes of our lab-based evaluation of the prototype.
Text Entry on Small DevicesBefore the widespread adoption of touchscreen smartphones, 12-key physicalkeypad phones were the most common text entry method on small devices. Predictive technologies interpreted the ambiguous keys (usually three or four letters per key) and suggested words. 1,2 This approach was shown to achieve speeds of approximately 10 words per minute (wpm) for novices and 20 to 25 wpm for experts in controlled studies. 3 In a previous study, we investigated this approach using a reduced number of keys for text entry on watches, but we implemented it on a touchscreen handheld device. 4 In theory, 12key ambiguous predictive text quality could be very high (over 90 percent accurate), but in reality, each key sequence could match many different words, and some of these sequences included pairs of common words that caused problems. (For example, on a standard phone keypad, "he" and "if" are typed with the same keys.) The early models of prediction were based on simple unigram dictionary models that suggested the most common word matching a sequence. Nowadays, phones have much more power and memory, so they can easily support more complex prediction models, greatly reducing the impact of ambiguity.Alternative approaches for input on small devices include handwriting, 5 fast but difficultto-learn chord keyboards, 6 and specialized alphabets. 7 Many domestic appliances, such as televisions and games, use a date-stamp-inspired method, where the user scrolls through the alphabet and picks letters from a 2D line or 3D grid. However, this has been shown to be a slow entry method. 8 Finally, gesture wordbased input techniques offer great benefits for mobile phones and can provide a fast and more relaxing input method than continuously tapping small on-screen buttons. 9 There is currently very little work focusing on text entry for smart watches. One such method is Zoomboard, where a full qwerty keyboard is shrunk to fill the smart-watch screen. 10 Users tap once to zoom into a keyboard area and a second time to select a letter from that area. Although shown to be good enough for input speeds up to 9 wpm, this interaction method lacks suggestion support and increases the number of interactions, because additional input is required for zooming. Furthermore, this
Although smart watches let users receive many forms of communication, there is usually no direct way of replying. The authors introduce an interaction design and an optimized alphabetic layout for smart-watch text entry and present an evaluation using a working prototype.