To address the challenges of polysemy, boundary ambiguity, and semantic-syntax coupling in Chinese Named Entity Recognition (NER), this paper proposes a Named Entity Recognition model based on Gated Character-Sentence Interaction. The model consists of five components: a character embedding layer, a sentence embedding layer, a character-sentence interaction extraction layer, a feature fusion layer, and a CRF decoding layer. Specifically, it first uses BERT to obtain context-sensitive character representations. It then designs a multi-view attention mechanism to divide sentence-level representations into four complementary semantic subspaces for global context modeling. Subsequently, it introduces a dimension reordering and convolution interaction module inspired by InteractE to extract deep interaction features between characters and sentences. Finally, it dynamically filters and fuses the dual-channel features through a gating mechanism and decodes the optimal label sequence using CRF. Experiments on four Chinese NER benchmarks—OntoNotes 4.0, MSRA, Weibo, and Resume—demonstrate that the model achieves F1 scores of 84.85%, 96.41%, 74.24%, and 97.11%, respectively, all of which exhibit highly competitive performance. Ablation studies confirm the effectiveness of the multi-view sentence representations and gated interaction. The research confirms that character-sentence interaction information can significantly enhance the performance of Chinese NER and possesses cross-domain robustness.