Noticias
June, 2023 – Cinthia Sánchez, a Ph.D. student from the Department of Computer Science at the University of Chile and the Millenium Institute Foundational Research on Data, presented the work “Cross-Lingual and Cross-Domain Crisis Classification for Low-Resource Scenarios” at The Seventeenth International Conference on Web and Social Media. This international conference brings together researchers from multiple disciplines with a common focus on online social networks.
The research, co-authored by IMFD researchers and academics from the Department of Computer Science at the University of Chile – Hernán Sarmiento, Andres Abeliuk, Jorge Pérez, and Barbara Poblete – focuses on utilizing social media data for disaster management through the automatic identification of crisis-related messages.
“In this work, we study the task of automatically classifying messages related to crisis events by leveraging labeled data across languages and domains,” explains Sánchez. The goal is to make use of labeled data in languages with abundant resources to classify messages in other languages that have fewer resources or different types of crisis situations.
Among the research findings, they determined that it is possible to use crisis data in English to classify the same and new types of crises in other languages, such as Spanish and Italian. “Overall, our work contributes to improving the data scarcity problem, which is crucial for multilingual crisis classification, particularly in mitigating situations from the onset of emergency events when time is of the essence,” highlights the researcher.
The 17th edition of the International Conference on Web and Social Media took place in Limassol, Cyprus, and featured a significant number of presentations and studies from different parts of the world. The conference has become a major meeting point for computational social science, showcasing articles, posters, and workshops based on network science, machine learning, computational linguistics, sociology, communication, and political science. The quality of the presentations has contributed to the rapid growth of the conference, which maintains a competitive acceptance rate of approximately 20% for full research papers published in the proceedings of the Association for the Advancement of Artificial Intelligence (AAAI).
