#Experimental code for Implicit Discourse Relation Classifier The code that replicates results from

Attapol T. Rutherford, Nianwen Xue (2014). Discovering Implicit Discourse Relations Through Brown Cluster Pair Representation and Coreference Patterns. Proceedings of the 14th Conference of the European Chapter of the Association for Computational Linguistics (EACL 2014)

Prerequisites

1. Stanford CoreNLP suite
2. Python2.7
3. sqlite3

Recipe for replicating the results

1. Parse PDTB data into jsons python2.7 tpl/pdtb_tools/parse_pdtb.py path/to/pdtb/* You will then get wsj_pdtb.jsons file, which contain each relation in a json string on each line.

2. Parse WSJ text data with StanfordCoreNLP into XML format. This process takes quite a bit of time because we are parsing a large number of sentences.

3. Convert the parses into json format python2.7 corenlp_xml2json.py path/to/xml_parses/* Then you will have json file for each wsj file, which gets put into the same folder as the xml parse folder.

4. Adding the data to the database and converting them db/init_db path/to/parse_folder. This will take quite a bit of time.

5. Make a data file python2.7 make_data_file.py Then you have training_set.csv and test_set.csv. The data are in tab-separated format. feature1 feature2\tlabel\twsj_section

6. Use machine learning package of your choice to run classifier. To follow the process described in the paper, we also need to exclude the features that appear less than 5 times and reweight the instances such that the positive and negative classes receive the same total weights.