Thursday, December 22, 2005

Text Entry Innovations – Tap, Type, Graph, Predict, or Say It

Making text entry easier and more efficient has obvious benefits to the mobile computing space. As Virtual Instrumentation moves from workstation to desktop stations to handheld devices, text entry becomes an important enabling technology.

Palm gave us the graffiti editor and the modern cell phone brought us Standard Ambiguous Code (SAC) which places “ABC” on the first key, “DEF” on the second key and so forth. The user makes multiple taps on a key (called multitap) to choose a letter. One tap on number 1 key brings up the letter “A”, two taps the letter “B”, and three taps the letter “C”.

Text entry systems reuse keys to create more input letters within a constrained space. Some systems use linguistic techniques to predict the word the user is entering and perhaps the one that will come next. Other systems create dictionaries (either user created or system created) to provide the words for prediction. Linguistic rules may not keep up with the ever changing language while dictionary systems require more memory to hold the required words.

Motorola’s iTap improves on this by allowing the user to hit the key once for each letter in the word to be entered and then using a dictionary to make out the correct word since each tap can represent one of three letters. T9 offers a demo on how multitap works and also provides a dictionary with updates to our ever growing language.

IBM’s “Alphaworks,” their nomenclature for emerging technologies, has created Shorthand-Aided Rapid Keyboarding. It uses “Sokgraphs” (Shorthand On Keyboard Graphs) which are letter tracings using a pen on a keyboard layout for either QWERTY or ATOMIK, that can be stored and used later. It takes some getting used to but it’s a natural progression from graffiti used by the Palm to outline each letter in a word you want to enter. To see it in action, check out the video clip here. Based on usage, the system can learn and start to predict words based on previously built words. IBM brands its stylus-based technology for text entry as Shapewriter, and offers a demo download here.

Eatoni makes predictive text entry systems for handheld devices. At first I thought the inventor was named Eatoni. Then I found their FAQ which indicates the name came from the six most common letters in the English language –E-a-t-0-n-i. Eatoni’s technology EQx stands for "Eatoni Qwerty x-column" which works in either 3 or 6 columns. EQx merges a QWERTY keyboard with a telephone keypad. Eatoni claims to extend the work of Scholes who invented the QWERTY keyboard by designing a keypad that predicts the word the user wants and avoids collisions in the keypad entry. Demos are available here.

Zi Corporation’s eZiTap uses a modified form of the multitap technique to perform word completion and prediction.

Senseboard is a Swedish company that offers a virtual keypad for text entry into mobile devices. Consisting of two hand-worn devices and a scanner, it detects the motions of the fingers as they type out a word on any surface. This eliminates the scrunching of the fingers trying to type on a tiny surface of keys.
Speech recognition seems an obvious technology and has been around for some time but has encountered numerous problems. In this review of a speech by Dr Gong Yi Fan the author divides the problem into three categories: performance, noise, and adaptiveness. In the area of performance, mobile devices are power constrained. Speech recognition can be processor intensive which can be power intensive. Background noise can also compromise speech recognitions’ capabilities. Finally, the speaker can also change his voice characteristics including pitch, volume, etc, challenging speech recognition’s ability to accurately translate the spoken word into textual entry.
If you are working with text entry systems, I would like to hear from you. You can reach me at hall.martin@ni.com.

Best regards,
Hall T. Martin