Derailment Forecasting Papers

Ever since my papers "Conversations Gone Awry" (ACL '18) and "Trouble on the Horizon" (EMNLP '19) introduced and formalized forecasting conversational derailment as an NLP task, the task has begun to be picked up by the broader NLP and Computational Social Science communities. This page is an effort to track new models for forecasting derailment from across the community.

If you know of a paper that you think should be included, or you are the author of a new paper or model for forecasting derailment, please contact me so I can add it to the list!

List of papers and models (sorted by publication date)

"Conversation Derailment Forecasting with Graph Convolutional Networks". Enas Altarawneh, Ammeta Agrawal, Michael Jenkin, and Manos Papagelis. arXiv preprint 2306.12982 (2023). [PDF]

"Conversation Modeling to Predict Derailment". Jiaqing Yuan and Munindar P. Singh. Proceedings of ICWSM 2023. [PDF]

"Multimodal Conversation Modelling for Topic Derailment Detection". Zhenhao Li, Marek Rei, and Lucia Specia. Findings of the Association for Computational Linguistics: EMNLP 2022. [PDF]

"Dynamic Forecasting of Conversation Derailment". Yova Kementchedjhieva and Anders Søgaard. Proceedings of EMNLP 2021. [PDF]

"Time Aspect in Making an Actionable Prediction of a Conversation Breakdown". Piotr Janiszewski, Mateusz Lango, and Jerzy Stefanowski. Proceedings of ECML PKDD 2021. [PDF]

"The Structure of Toxic Conversations on Twitter". Martin Saveski, Brandon Roy, and Deb Roy. Proceedings of WWW 2021. [PDF]

"On Predicting Behavioral Deterioration in Online Discussion Forums". Jean Marie Tshimula, Belkacem Chikhaoui, and Shengrui Wang. Proceedings of ASONAM 2020. [PDF]

"Preemptive Toxic Language Detection in Wikipedia Comments Using Thread-Level Context". Mladen Karan and Jan Šnajder. Proceedings of ALW 2019. [PDF]