This dataset is a multimodal public opinion data of 9.5 Sichuan Luding topic netizens collected and annotated with emotional types on Weibo social media platform.[Collecting] Retrieved search results using the keyword "Luding Earthquake" from September 5th to 15th, 2022, and obtained 35094 original Weibo posts.[Processing] The original dataset is filtered to obtain a Weibo dataset containing videos. Then, short data with a video duration of less than 3 seconds is further removed, resulting in 6270 Weibo data containing videos of sufficient length. Due to the phenomenon of media using the same text or video to publish Weibo multiple times through their account matrix, data was deduplicated by comparing text and video files, and irrelevant or meaningless content was removed. Finally, 2703 Weibo data were obtained after cleaning, each data being a text video pair.[Emotional Labeling] Six master's students majoring in Library, Information and Archive Management were recruited to annotate data with emotions. The members annotated the data with emotional labels based on the text content and video content (divided into three categories: positive, negative, or neutral). The expression of concerns and hopes for emergency rescue and disaster relief, as well as the content of search and rescue activities and successful evasions, reflects the positive emotions of netizens; The expression of disaster situations leading to loss, fear, helplessness, and concern reflects negative emotions; There is no obvious emotional tendency in relevant factual reports, knowledge popularization, refutation of rumors, praying for blessings, and ridicule, which belong to neutral emotions.The kappa coefficient of the annotation result is 0.812, which passes the annotation consistency test. Inconsistent annotation labels were thoroughly discussed and determined by team members, resulting in a consistent annotation dataset.