<p>Event extraction aims to identify event triggers and extract their corresponding arguments from unstructured text. A key challenge is that event elements are expressed through both contextual semantics and structural dependencies, while existing methods often underuse the interaction between sentence-level syntax and schema-level event knowledge. In this paper, we propose a novel <b>Structure-Guided Representation Enhancement</b> (SGRE) framework that aligns event template structures with syntactic graphs through cross-structural alignment. SGRE represents event schemas as structured prior knowledge rather than isolated labels, and converts them into token-level guidance for trigger and argument prediction. Specifically, an <i>Event Template Encoder</i> models event-type and role-type structures, a <i>Syntactic Graph Encoder</i> captures dependency-based sentence structures, and a <i>Structural Alignment Network</i> with adaptive gating integrates these structural signals with contextual representations. The main innovation lies in bridging bottom-up syntactic evidence and top-down schema constraints through attention-based alignment, producing more discriminative representations for event extraction. Experimental results on GENIA11 and FewFC show that SGRE achieves competitive results against representative baselines on both trigger and argument extraction. Ablation and visualization analyses further confirm the effectiveness of the proposed structural components.</p>