Text-entry systems on special purpose keyboards require overloading multiple characters on single keys to permit entering English text.  Such systems can be cumbersome to use because multiple keypresses are necessary to distinguish the actual intended character.

We have developed a new technology which uses statistical and grammatical constraints and algorithmic techniques to resolve the ambiguity inherent in keyboard overloading with high accuracy.
Applications include:

Telephone keypad text entry
Standard telephone keypads are labeled with three letters on each key, creating an ambiguity as to which character was intended, which must be resolved for unambiguous text entry. Existing systems all use pairs of keypresses to spell out single letters, but are extremely cumbersome and frustrating to use. Our reconstruction engine correctly identifies up to 99% of the characters in typical text, eliminating the need for multiple keypresses in many applications.

Minimizing motion on QWERTY keyboards
To speed text entry or minimize finger motion on conventional QWERTY keyboards, our techniques can be used to resolve ambiguity so that a typist need not move their fingers off the home row of the keyboard.

Customized keyboards for text entry
Special-purpose keyboards can be made more efficient via our prediction techniques, such as one-handed chord or half-QWERTY keyboards, and court stenography keyboards.

Keyboards for the disabled
Severely disabled people often lack sufficient motor control to type on a keyboard of more than 6-10 keys. Further, each stroke typically requires non-trivial concentration and effort, so minimizing the number of keystrokes is critically important to creating a usable system. We propose using heavily overloaded keyboards, with the ambiguity being resolved using our methods, thus significantly reducing the effort needed to enter text.

Error Correction of Scanned Text
Both handwriting and optical character recognition (OCR) systems suffer from substantial rates of character misrecognition. Substantial error correction can be obtained by using our reconstruction engine to exploiting grammatical and statistical features of English.