With the rapid development of Chinese internet neologisms, many new words exhibit polysemy and strong context dependence, posing challenges to traditional semantic analysis methods. To address this issue, we propose an innovative method for the interpretation of Chinese internet neologisms—Multi-CoT Enhanced New Word Understanding (MCot-NWU). This method combines multi-dimensional collaborative reasoning and BERT perplexity measurement to effectively evaluate and select the optimal interpretation. MCot-NWU features three key innovations: (1) a multi-dimensional collaborative reasoning mechanism that breaks the limitations of traditional single-chain reasoning by constructing parallel reasoning channels across six semantic dimensions (homophones/near-homophones, symbols/numbers, etc.), enabling collaborative semantic disambiguation; (2) a hierarchical fusion architecture with a dual-layer design, featuring customized reasoning chains for different word types and dynamic weight allocation via BERT-based perplexity metrics; and (3) a composite evaluation system that upgrades traditional semantic similarity calculations to a three-dimensional model integrating contextual fit, semantic reasonableness, and cultural compatibility. The system architecture includes a front-end classifier for processing six types of new words, a mid-end parallel multi-chain reasoning engine, and a rear-end fusion module based on Attention-BERT. Experimental results demonstrate that MCot-NWU achieves exceptional performance, particularly in categories such as “homophones/synonyms” and “pictograms/actions,” significantly outperforming single-chain and multi-chain models. By reducing conflicts and redundancy in the reasoning process, MCot-NWU enhances the accuracy and robustness of semantic interpretation, providing a robust solution for the dynamic and complex nature of Chinese internet neologisms.