PhD Thesis · Federico Pardo
This thesis addresses the gap between algorithmic modeling and teaching practice by proposing a computational architecture for synchronous classroom analysis. By fusing paralinguistic features and multilingual semantic analysis, the research demonstrates that audio contains stratified levels of pedagogical information capable of activating teacher reflection processes through objective and transparent metrics.
The following peer-reviewed publications conform the core basis of the research developed during this PhD.
Combined text (BERT) and audio features to classify different types of teacher interventions in classrooms.
View Publication →Systematic review of 82 studies (2014-2024) on using audio processing in educational research.
View Publication →Explored techniques to make multimodal AI models generalize better across different classroom settings.
View Publication →