GitHub - jpate/prosodicParsing: different kinds of HMMs to use for incorporating prosody into basic parsing

jpate / prosodicParsing Public

Notifications You must be signed in to change notification settings
Fork 3
Star 2

different kinds of HMMs to use for incorporating prosody into basic parsing

homepages.inf.ed.ac.uk/s0930006

Notifications

Name		Name	Last commit message	Last commit date
Latest commit History 94 Commits
lib		lib
project		project
src/main/scala		src/main/scala
target		target
testdata		testdata
README		README

Repository files navigation

Here are some Hidden Markov Models (HMMs).

PlainHMM is the usual HMM:

 q_1  --->  q_2 --->  q_3 ....
  |          |         |
  |          |         |
  v          v         v
 o_1        o_2       o_3 ...


TwoOutputHMM has one hidden stream and two output streams:

 a_1        a_2       a_3 ...
  ^          ^         ^
  |          |         |
  |          |         |
 q_1  --->  q_2 --->  q_3 ...
  |          |         |
  |          |         |
  v          v         v
 o_1        o_2       o_3 ...


CoupledHMM has two hidden streams and two output streams:

 a_1        a_2        a_3 ...
  ^          ^          ^
  |          |          |
  |          |          |
 p_1  -->_  p_2  -->_  p_3 ...
     \   /|     \   /| 
      \ /        \ /
       X          X
      / \        / \
     /  _\|     /  _\| 
 q_1  -->   q_2  -->   q_3 ...
  |          |          |
  |          |          |
  v          v          v
 o_1        o_2        o_3 ...



This uses a modified version of mallet (included in the lib directory) whose source can be obtained
from:

https://bitbucket.org/jpate/mallet