End-to-End Deep Reinforcement Learning for Conversation Disentanglement
Published in Proceedings of the 37th AAAI Conference on Artificial Intelligence (AAAI 2023), 2023
Formulates conversation disentanglement as a sequential decision-making problem and trains an end-to-end deep RL agent to cluster interleaved multi-party chat messages into coherent threads. The approach outperforms prior pairwise-classification pipelines on standard IRC benchmarks.
Recommended citation: Bhukar, K., Kumar, H., Raghu, D., Gupta, A. (2023). "End-to-End Deep Reinforcement Learning for Conversation Disentanglement." Proceedings of the 37th AAAI Conference on Artificial Intelligence.
Download Paper