rss
 
comment(s)

archives
J|F|M|A|M|J|J|A|S|O|N|D
(20##) 10 9 8 7 6 5 4 3 2 1 0 <
 
DesktopWeb FormText   hidden markov modelsWed, 06 Oct 2004 23:23:44 GMT # 

in some recent article i mentioned 'hidden markov models' ... and admitted that the name was familiar, but i had forgotten the concept. so i spent most of today brushing up. the 'hello world' of HMM is to feed it a body of text from a specific author and then use that to generate a new sentence that has a likely probability of occuring. ported some C code over and ran it against the transcript of the presidential debates. it was not as funny as i had hoped for. if i told it to start with the word 'osama', then it did generate 'osama bin laden'. could also get Bush's corpus to say 'it's hard work' by making it start with the word 'it's'. HMMs can be applied to speech recognition, natural language processing, music modeling, and spell checking. now i'm trying to come up with a different project idea in which it would make sense to use an HMM. NOTE the use of the word 'corpus' above ... i've been reading too many academic books about AI ...