Thread: handwriting: why up/downcase distinction?
View Single Post
Karel Jansens's Avatar
Karel Jansens is offline Karel Jansens
2008-05-02 , 13:11
Posts: 3,220 | Thanked: 326 times | Joined on Oct 2005 @ "Almost there!" (Monte Christo, Count of)
#27
Originally Posted by ragnar View Post
HWR is rather tricky also because of the intellectual property right issues: there are _many_ patents held by different companies on various features, and they make it naturally tricky for others to keep on improving the overall experience - you'll just start breaking the patents.

(You can look it the other way round, if it would be easy to do, then there would almost certainly be already good open source alternatives available, since there is is the need...)
I've lost the bookmark, but a few years ago someone wrote on one of the Linux HWR projects that he had developed a natural handwriting algorithm which he would like to "inject" into that particular project (I'm thinking it was X-stroke, but it could have been any of the others). So there was at least one "free" algorithm around.

Basically, any HWR that's worth its musterd is going to use a neural network for recognition, coupled with good dictionaries. The only IP is in the preconfigured NN's "knowledge", nobody can stop you from setting up and distributing your own training.
__________________
Watch out Nokia, Pandora's box has opened (sorta)...
I do love explaining cryptic sigs, but for the impatient: http://www.openpandora.org/