ChantalvanSon / unsharedtask2016

Introduction

This repository provides the data and code used for the work described in:

C. van Son, T. Caselli, A. Fokkens, I. Maks, R. Morante, L. Aroyo, and P. Vossen (2016). "Unshared Task at the 3rd Workshop on Argument Mining: perspective based local agreement and disagreement in online debate." In Proceedings of the 3rd Workshop on Argument Mining, Berlin, Germany.

This work was conducted in the context of the Unshared Task of the 3rd Workshop on Argument Mining. The supplementary data for the Unshared Task was provided here. For the annotations described in our paper, we used the development sets of Variants C (editorial articles) and D (discussions). The annotations were done using the Content Annotation Tool (CAT) [2].

Data

The folder data contains the texts that were annotated as described in our paper. The texts have been pre-processed in several steps, resulting in different formats.

1. editorial_articles: this folder contains a set of 8 editorial articles (the development set of Variant D of the Unshared Task). More specifically, it contains 4 subfolders resulting from several pre-processing steps:

• original: contains the original plain text files as provided for the Unshared Task.
• stripped: contains plain text files containing the title and the paragraphs of the editorial article. Created by removing 'meaningless' new lines and meta-data from the original text files.
• NAF-tokenized: contains tokenized texts in NAF format. The English version of the ixa-pipe-tok module was used for the tokenization and sentence splitting.
• CAT-tokenized: contains tokenized texts in CAT input format. They were created using the naf2cat.py script.

2. discussions: this folder contains a corresponding set of 8 discussions about the editorial articles (as originally provided for the Unshared Task). In total, the discussions contain 62 comments.
3. Task1-annotations: this folder contains the documents as they were annotated (including both the raw input and the annotated versions).

• CAT-input-format: contains files consisting of a comment (first sentence in CAT format) and the corresponding editorial article (the remaining sentences). They were created using the write_cat_files.py script
• CAT-Round-1: contains the annotations of Round 1 in CAT XML format.
• CAT-Round-2: contains the annotations of Round 2 in CAT XML format.

Code & Replication

The documents were pre-processed for the annotations using the following scripts and modules in this order:

• 1-write_stripped_articles.py
• 2-ixa-pipe-tok
• 3-naf2cat.py
• 4-write_cat_files.py

Some of these scripts make use of the following two general modules:

• cat-information.py: contains general functions for getting information from CAT XML files
• naf-information.py: contains general functions for getting information from NAF files

To get the distributions of the annotations as described in our paper, the following script was used:

• get_distribution_annotations.py

References

[1] C. van Son, T. Caselli, A. Fokkens, I. Maks, R. Morante, L. Aroyo, and P. Vossen (2016). "Unshared Task at the 3rd Workshop on Argument Mining: perspective based local agreement and disagreement in online debate." In Proceedings of the 3rd Workshop on Argument Mining, Berlin, Germany.

[2] V. B. Lenzi, G. Moretti, and R. Sprugnoli (2012). "CAT: the CELCT Annotation Tool". In Proceedings of the 8th International Conference on Language Resources and Evaluation (LREC 2012), Istanbul, Turkey. (pdf)