Abstract: 【Purpose/significance】This thesis investigates the communication rules and patterns of public opinion in public emergen⁃ cies on short video platforms. Its purpose is to give insights about public opinion evolution in new media environments and improve government response strategies.【Method/process】We collect comments from BiliBili platform as a case study to using 2021 Henan ex⁃ treme rainfall disaster. Then built sentiment analysis model using sentiment dictionaries and applied the methods such as topic cluster⁃ ing, sentiment polarity analysis and temporal evolution analysis.【Result/conclusion】Comment topics focused on four areas: Disaster descriptions, mutual aid information, emotional expression, and rescue operations. They were mostly positive comments. The evolution followed four stages: Emotional catharsis - Mixed emotions - Rational thinking - Positive restoration.【Innovation/limitation】With the context of public emergencies, we discussed public psychological mechanisms and social response patterns from multi-angle perspec⁃ tives of emotional dynamics, social identity and social capital.