This paper presents TrendScope, a temporal hypergraph-based framework for detecting emergent food trends from social media. Traditional models such as keyword bursts and pairwise graphs are limited in capturing the high-order, multi-entity relationships that characterize trend emergence. TrendScope models each social post as a hyperedge connecting heterogeneous entities—users, food items, hashtags, locations, and temporal bins—thus preserving rich semantic interactions. A sequence of time-indexed hypergraph snapshots is constructed and encoded using hypergraph neural networks. Temporal aggregation is performed via attention-based or recurrent mechanisms to learn expressive, time-aware node embeddings. A composite scoring function, integrating temporal frequency shifts, structural centrality changes, and semantic embedding divergence, ranks candidate food entities by trend significance. Evaluations on real Reddit data and a synthetic benchmark demonstrate superior performance over graph-based and static hypergraph baselines in both trend accuracy and interpretability. The framework is scalable, interpretable, and applicable to industrial domains such as menu optimization, targeted marketing, and supply chain forecasting.