A Refined Discourse Parser for CoNLL 2015 Shared Task

A Refined End-to-End Discourse Parser participated in CoNLL 2015 Shared Task is described in the paper A Refined End-to-End Discourse Parser by Jianxiang Wang and Man Lan.

Requirements

• ETE
• MALLET
• Python >= 2.7

Usage

First, change the values of the following two variables in config.py:

1. CWD: current working directory (absolute path).
2. MALLET_PATH: mallet bin path (absolute path).

Such as:

CWD = "/Users/XXX/Documents/conll2015_discourse/"
MALLET_PATH = "/Users/XXX/Documents/mallet"

Then, run the parser using the command in terminal as required in the CoNLL 2015 Shared Task Official Blog:

python $input_dataset $input_run $output_dir

• $input_dataset: the folder of the dataset to parse. The folder structure:

  $input_dataset/pdtb-parses.json

  $input_dataset/raw/...

• $input_run: the folder that contains the model file or other resources. Here, simply set it to 'none' in our parser.

• $output_dir: the folder that the parser will output 'output.json' to.

For example:

python parser.py data/conll15-st-03-04-15-dev none data

The parser will take the dataset under the 'data/conll15-st-03-04-15-dev' folder as input and generate a 'output.json' which contains the discourse relations parsed by the parser under the 'data' folder.