
Samawel Jaballi has been invited to participate in ICCCI 2025 (17th International Conference on Computational Collective Intelligence - Core B), to be held from November 12 to 15, 2025, in Ho Chi Minh (Vietnam). His paper entitled: “A Study in UNESCO's Audio Messages for Rapid Health Response” has been accepted.
ICCCI 2025 is an international conference, jointly organized by the University of Wroclaw (Poland) and VNU-HCM (Vietnam).
During these exchanges, in the field of collective computational intelligence and its applications, topics related to the processing of data and knowledge from autonomous sources will be addressed, in particular: collective processing, knowledge integration, data integration, group decision-making, multi-criteria decision-making, consensus computing and social networks.
Samawel's work focuses on keyword detection in continuous, multilingual audio streams, in a health crisis context. The proposed approach combines enriched acoustic descriptors (MFCC and prosodic signals) with reweighted semantic representations based on M-BERT embeddings, integrated into optimized ResNet models. The study shows that our architecture, recalibrated by dynamic residual connections, outperforms deeper models while limiting the overlearning problems encountered on classical benchmarking models.