An analyst or a planner seeking a rich, deep understanding of an emergent situation today is faced with a paradox – multimodal, multilingual real-time information about most emergent situations is freely available but the sheer volume and diversity of such information make the task of understanding a specific situation or finding relevant information an immensely challenging one. To remedy this situation, the Generating Alternative Interpretations for Analysis (GAIA) team at DARPA AIDA program aims for automated solutions that provide an integrated, comprehensive, nuanced, and timely view of emerging events, situations, and trends of interest. GAIA focuses on developing a multi-hypothesis semantic engine that embodies a novel synthesis of new and existing technologies in multimodal knowledge extraction, semantic integration, knowledge graph generation, and inference. In the past year, the GAIA team has developed an end-to-end knowledge extraction, grounding, inference, clustering and hypothesis generation system that covers all languages, data modalities and knowledge element types defined in AIDA ontologies. We participated in the evaluations of all tasks within TA1, TA2, and TA3. The system incorporates a number of impactful and fresh research innovations.
Tongtao Zhang , Ananya Subburathinam , Ge Shi , Lifu Huang, Di Lu, Xiaoman Pan, Manling Li, Boliang Zhang, Qingyun Wang, Spencer Whitehead, Heng Ji, Alireza Zareian, Hassan Akbari, Brian Chen, Ruiqi Zhong, Steven Shao, Emily Allaway, Shih-Fu Chang, Kathleen McKeown, Dongyu Li, Xin Huang, Kexuan Sun, Xujun Peng, Ryan Gabbard, Marjorie Freedman, Mayank Kejriwal, Ram Nevatia, Pedro Szekely, T.K. Satish Kumar, Ali Sadeghian, Giacomo Bergami, Sourav Dutta, Miguel Rodriguez, Daisy Zhe Wang (Information Sciences Institute; Columbia University; Rensselaer Polytechnic Institute; University of Florida; University of Southern California)