Touch-Tone Telecommunication Aid for the Hearing-Impaired
This award-winning device, which I invented for my Master's Thesis Project at MIT, creates readable text when people type sentences using the Touch-Tone keypad, using the redundancy of language to decide which of the three letters on a given key was the intended one. The device I built was intended to serve the needs of deaf telephone users (I'd had a deaf roommate and understood the problem he faced quite well), by enabling them to simply and rapidly receive text from communicative partners who had no special equipment at their end of a phone connection. It was a groundbreaking device, worked very well, and won the National Student Design Contest from the Rehabilitation Engineering Society of North America.
The device used a hybrid algorithm that incorporated a dictionary-lookup method and a Markov-chaining technique. This hybrid algorithm allowed the system to perform well in both the routine situations where the text strings were in the dictionary yet also come up with readable words that weren't in the dictionary. Today, with memory cheap and plentiful, most algorithms for this purpose can do without the probabilistic element.
This work was done years before the introduction and spread of the cellular phone, and I cannot say that I was anticipating the rise of texting when I performed this research and implementation. In fact, it was seeing the needs of my deaf roommate and realizing that the Touch-Tone keypad could alleviate some of his frustrations with the telephone (and partly the introduction of chipsets that could decode Touch-Tones in consumer devices) that got me started on this project. My work was released into the public domain, and it's a mystery to me and legal experts alike how the T9 algorithm was ever granted a patent, and even more of a mystery that major corporations licensed it. I'll happily testify, if anyone wants to be free of the legal shackles around text entry on ambiguous keypads.