MANILA26: SIGIR 2026 Tutorial on Information Retrieval for Climate Change Impact

Le résumé fourni par la source

This three-hour tutorial uses climate change evidence synthesis as a high-stakes ''stress test'' for state-of-the-art IR and AI methods. Participants will explore why this domain, which is characterized by high levels of interdisciplinarity, multi-modal data (geospatial and time-series), and a lack of controlled vocabularies, challenges current state-of-the-art systems and active learning protocols. Through an architectural critique of agentic RAG systems and a deep dive into the ''AMOC experiment,'' the tutorial distinguishes between automated literature review and expert scientific assessment. Participants will learn to diagnose specific retrieval failure modes, such as geographic coverage bias and the ''tacit knowledge boundary,'' where human expertise remains essential. The tutorial concludes by framing a new research agenda: building unified, verifiable evidence pipelines that integrate bibliographic, geospatial, and multi-modal retrieval with rigorous provenance tracking.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.