[Tutor] need some help for program!! [More hidden markov models in Python]

[Danny Yoo]

> Again, the page at:
>
>     http://datamining.anu.edu.au/software/febrl/febrldoc/node7.html
>
> has an example that should help clarify why HMM's are appropriate for
> your problem.

Hi Sachin,

Argh, darn it!  I can't help myself but look into your question some more.
Why did you have to ask a question on such an interesting topic?  *grin* I
have to control my curiosity or else I'll end up writing this program too.


There's a great source of information on Hidden Markov Models in the
source code of Logilab's 'hmm' module:

    http://www.logilab.org/hmm/

This is a Python module that implements HMM's, and is excellent.  You may
need to install an extension called Numeric Python to get it to work, but
it's definitely worth it:

    http://www.pfdubois.com/numpy/


There are some real gems in there, not just the module itself, but in the
test cases they've embedded in the source code.  In particular, if we take
a close look at the logilab.hmm.hmm.test2() function, we'll see that they
train a Hidden Markov Model to recognize French sentences (and evaluate
their likelihood!).  If you read and understand their test2() function,
you should be able to apply that knowledge toward your problem.


Another good source of information on HMM's includes the classic article,
"A Tutorial on Hidden Markov Models and Selected Applications in Speech
Recognition", by Lawrence R. Rabiner.  Here's a link to the paper:

    http://www.cc.gatech.edu/ccg/paper_of_week/Rabiner-Markov.pdf

I just read the first few pages, and my own fuzzy ideas about HMM's just
became a lot clearer.


Anyway, I must stop myself now, or else I won't be able to sleep.  This is
why I hate seeing homework questions...

Good luck to you.