UNSC-RST: Rhetorical Strategies in the UN Security Council

Dataset, Amendment for RST annotation guidelines and code for experiments as described in the paper:
Karolina Zaczynska, and Manfred Stede, "Rhetorical Strategies in the UN Security Council: Rhetorical Structure Theory and Conflicts" Proceedings of the 25th Meeting of the Special Interest Group on Discourse and Dialogue (SIGDIAL 2024). Kyoto, Japan, 2024

This git-repository includes a new dataset with RST annotations of 82 diplomatic speeches given in the UN Security Council (UNSC-RST). For the RST annotations we use a modified version of the guidelines by Stede et al. 2014.

For the analysis of rhetorical strategies used by diplomats to express (verbal) Conflicts, we aligned the RST and existing Conflict annotations (see also our previous paper on Conflicts). We look at both the RST-relation distribution and the RST tree structure in combination with Conflicts in the speeches. In preliminary analyses we already see patterns that are characteristic for particular topics and countries.

Note: The dataset and code are the one used for the SIGDIAL paper. An updated version can be found in our UP Multilayer UNSC Corpus (UMUC) repository: https://github.com/discourse-lab/UMUC

Dataset short description

Note: A more detailed description of the dataset files is in data/data_readme.md.

\data folder contains:

• \output directory with computed Nuclearity Mass values per speech (nuclearity_mass_speech.csv) or paragraph (nuclearity_mass_para.csv) and per paragraph with paragraph-based RST-subtree values (main_conflicts_aligned_paragraohid.csv).
• 07_rst RST annotations (retrieved from RSTweb tool) in rs3-format
• main_conflicts.csv UNSC speechs with Conflicts annotations per EDU (taken from https://github.com/linatal/UNSCon)
• main_conflicts_aligned.csv UNSC speechs with Conflicts annotations with RST annotations per EDU aligned

Code

To work with skripts, create a Virtual Environment using python 3.10 (not tested with other versions):
python3.10 -m venv myenv
Download SpaCy Language Model:
python -m spacy download en_core_web_lg
Install the Required Packages:
pip install -r requirements.txt