This work studies how movie genre and user-specific behavior shape spoiler likelihood. It combines dynamically modeled review history with genre-aware aggregation and a mixture-of-experts architecture, allowing the detector to capture both recurring author bias and genre-dependent language patterns. The approach improves spoiler detection across benchmark review datasets and was published at ECML PKDD 2025.
Genre structure and user history expose spoiler patterns that text-only classifiers tend to miss.