Named Entity Recognition (NER) is crucial in NLP for identifying entities in unstructured text. Span-based models are effective but often ignore semantic and structural dependencies between candidate and entity spans. To address this, we propose the Tri-Affine Attention-Enhanced Spanning Network. This model integrates a tri-affine attention module with a sophisticated spanning encoder, which leverages both multi-head self-attention and cross-attention mechanisms to enhance inter-span interaction and contextual encoding. The tri-affine attention module is used to accurately model the complex interactions between candidate spans and entity spans, thereby reducing boundary prediction errors, while the improved span encoder effectively improves the boundary-awareness of the model by capturing long-range dependencies and semantic interactions across spans. Experimental results across multiple benchmark datasets demonstrate that the proposed method consistently achieves superior performance compared to previously established state-of-the-art approaches.