This project intends to tag the dialog act given a discourse.

The corpus we used is from loria.fr. The corpus is used for a French Learnig dialog system. You can check corpus/sample to have a feeling about how the data looks like.

To use datagger, you need to install:

• NLTK: NLTK is a leading platform for building Python programs to work with human language data. http://nltk.org/
• CRF++: CRF++ is a simple, customizable, and open source implementation of Conditional Random Fields (CRFs) for segmenting/labeling sequential data. http://crfpp.googlecode.com/svn/trunk/doc/index.html
• scikit-learn: sickit-learn is python package for machine learning. http://scikit-learn.org/stable/
• KEA: KEA is a French tokenzier. https://github.com/boudinfl/kea

You can simply run

python2 Corpusbuilder --help
usage: CorpusBuilder.py [-h] [--cv] [--path PATH] corpuspath {crf,sample}
positional arguments:
 corpuspath The path to the corpus
 {crf,sample} select what kind of corpus to generate. sample: to generate
 unlabled data. crf: to build training and test data for crf
optional arguments:
 -h, --help show this help message and exit
 --cv build cross_validation corpus
 --path PATH place to put the generated corpus data

to check the usage of the script.

After you genrate the CRF++ data format. Simply use

crf_learn template train_data model`

you will get a model file. Use this model to tag test data, run

crf_test model test_data`

To evaluate our result, you can use the scipt from conll phrase recognition task. http://www.cnts.ua.ac.be/conll2000/chunking/output.html To evaluate our result, run:

perl conlleval.pl -r -d '\t'` < result_you_got

` To check more information of this project, read our report.