Dataset for the paper "How to disagree well: Investigating the dispute tactics used on Wikipedia" by De Kock at al., published in EMNLP 2022.
Dataset consists of a list of dicts, each representing a disagreement. Each disagreement dict has 4 keys: conv_id, utterances, split and escalation_label. Split refers to the split used by us, whereas escalation_label is a conversation-level constructiveness label from WikiDisputes (De Kock and Vlachos, 2021).
The utterances are also a list of dicts, each containing text, username, rebuttal_labels and coordination_labels.