DQNC2S: DQN-Based Cross-Stream Crisis Event Summarizer

Summarizing multiple disaster-relevant data streams simultaneously is particularly challenging as existing Retrieve &Re-ranking strategies suffer from the inherent redundancy of multi-stream data and limited scalability in a multi-query setting. This work proposes an online approach to crisis timeline generation based on weak annotation with Deep Q-Networks (DQNs). It selects on-the-fly the relevant pieces of text without requiring human annotations or content re-ranking. This makes the inference time independent of the number of input queries. The proposed approach also incorporates a redundancy filter into the reward function to handle cross-stream content overlaps effectively. The ROUGE and BERTScore results achieved on the CrisisFACTS 2022 benchmark are better than those of the best-performing models.