The goal of this thesis is the integration of Stream Reasoning techniques for highly-efficient reasoning in the presence of noise and uncertainty into a system expressive enough to handle contemporary Complex Event Recognition applications effectively. We focus on logic-based approaches which have proven both expressive and scalable, such as the Run-Time Event Calculus, and systems based on probabilistic logic programming, such as Prob-EC. We aim at developing a neuro-symbolic framework by integrating both approaches, while maintaining high efficiency, expressive power and robustness to uncertainty.