Rapid and updated disaster-damage information is crucial for making time-sensitive decisions. Social-media networks facilitate the dissemination of emergency information, which has been investigated as a useful tool to derive and analyze damage information. Here we develop a framework for rapid damage classification and recovery monitoring for urban floods that makes use of social-media big data (specifically, Weibo) by using an example that happened in Chongqing, China, from 18 to 20 August 2020. We first investigated the damage categories, including physical damage and emotional damage based on machine-learning classification algorithms. Then, we adopted a statistical model to measure the influencing factors of emotional responses to victims. Finally, we examined the ability of recovery based on keyword frequency representing flood destructive scenarios. We find that social-media data can be used to measure the level of damage and the state of the recovery. In the case we studied, regions with more physical damage tended to express more negative emotions, and government employees tended to convey positive information to reduce public panic when disaster occurs, while students were more likely to express negative emotions. The framework presented in this article provides a feasible way to obtain on-site damage information prior to conducting comprehensive surveys in disaster-stricken areas, which is helpful for real-time managing strategies for regional flood mitigation, as well managing as the sustainable development of flood-prone areas.