Workshop Proceedings of the 19th International AAAI Conference on Web and Social Media

Workshop: Workshop on Data for the Wellbeing of Most Vulnerable

DOI: 10.36190/2025.03

Published: 2025-06-05
A Multimodal TikTok Dataset of Ecuador's 2024 Political Crisis and Organized Crime Discourse
Gabriela Pinto, Emilio Ferrara

We present EcuadorCrisisTikTok, a large-scale, multimodal dataset comprising 51,479 TikTok videos that document the 2024 political crisis in Ecuador. Spanning content published between December 31, 2023, and March 4, 2024, the dataset includes rich video metadata, Whisper-generated transcripts, sentiment annotations, and LLM-generated summaries. Videos were collected using a targeted set of crisis-related hashtags curated from news reports and trending discourse. To demonstrate the dataset's utility, we conduct three example analyses: multilingual sentiment classification, video clustering using TimeSformer embeddings and UMAP, and topic modeling of transcripts. This dataset offers a valuable resource for research in crisis informatics, political communication, and misinformation in Latin America. In accordance with platform policies, only anonymized metadata and transcripts are publicly released. The dataset is available at: https://github.com/gabbypinto/EcuadorCrisisTikTok.