G-Tek wrote:
Has anyone seen or worked with anything like this in the past?
Not T9, but been playing with something similar. Kinda started with a Wiktionary-download, gives you a decent dictionary. Once that was in place, started fetching websites to do a statistical analysis. Read a few sites in a loop (of the correct target-audience) and mark how often a word follows the previous word. Thus resulted in a crash of my machine :) Filtered out all noise-words for the analysis, as there's a lot of words that might follow the word "this". The remark on the target-audience would be important because "this" often has a different meaning on the CodeProject-site than in normal conversation. ..and that's when I stopped playing with it, my harddisk isn't that large. You wouldn't need the complete resultset to give decent options as a prediction; only the top 5000 or so. Still, sounds like a lot of work, and the method that I described here is probably one of the less efficient ways to do it.
I are Troll :suss:
