MAXENTROPY DA CLASSIFIER FOR SAMSUNG DIALOG CORPUS
----------------------

This folder contains the code necessary for DA classifier and folders for evaluating it.
This implementation is based on 
S. Bangalore, G. Di Fabbrizio, and A. Stent, "Learning the structure of task-driven human–human dialogs," Audio, Speech, and Language Processing, IEEE Transactions on, vol. 16, pp. 1249-1259, 2008.


LICENSE
-------

The code is written in Python and is tested on 64bit Ubuntu Linux machine (4 cores and
6.4GB RAM) using Python 2.7. It is not tested on other platforms and might or might not
work (as Python itself is platform independent, the chances are you might be able to
get it worked as it is on other platforms too, but some minor modifications might be required.)
 
The code has several dependencies and these must be downloaded and installed first on
your system. 

DEPENDENCIES
-------------

1. Maximum Entropy Modeling Toolkit for Python and C++
   http://homepages.inf.ed.ac.uk/lzhang10/maxent_toolkit.html

2. NLTK: Natural Language Tool Kit in Python.
   http://www.nltk.org/


GETTING STARTED
---------------

To train and test using the MaxEntropy classifier use the following script diact_classifier.py .
In this script, evaluation using cross validation is performed

ss_dialog :  Samsung corpus for DA classification
map_test : DA classification result on Map corpus

CONTACT
-------
For questions, comments etc. contact Yeonchan Ahn.
[email protected]