Abstract Argumentation and Machine Learning for Process Event Analysis

A new approach combines abstract argumentation frameworks (AAF) with machine learning to analyze low-level process event streams. The method addresses the interpretation problem where trace events must be mapped to reference business activities. By framing interpretation as an acceptance problem within AAF, it analyzes plausible event interpretations and offers explanations for conflicts with prior knowledge. In highly uncertain settings, a sequence-tagging model trained to suggest probable candidates improves efficiency. This hybrid reasoning-based and data-driven technique reduces computational load and enhances informativeness.