Consistency of EU and Chinese Textile Fiber Content Labelling Regulations Based on Automated Auditing

The textual structure and terminology of Chinese and European textile fiber content labeling regulations differ significantly, thus it is challenging to conduct an efficient and accurate consistency check through traditional means of human comparison. Therefore, this paper proposes an automated auditing approach that relies on the semantic comparison of regulatory texts comparing the regulations across languages. It constructs a bilingual semantic knowledge graph of the Chinese and European regulations, utilizing a cross-lingual BERT (Bidirectional Encoder Representations from Transformers) embedding model to vectorize regulatory clauses, supplemented by a graph matching algorithm to calculate the degree of semantic similarity between clauses. These processes facilitate the automatic identification of regulatory consistency and enable quantitative comparison across regulatory systems. Experimental results indicate that the methodology yields F1 scores of 96.3% and 96.8% respectively in the clause segmentation module for Chinese and EU regulations, and 93.3% and 94.0% in the terminology extraction module. In producing semantic nodes, this paper achieves a highest accuracy of 95.8%, demonstrating a strong performance in cross-lingual parsing of regulatory texts. The system also maintained a stable alignment rate within a range of 0.85 to 0.91 in the bilingual mapping of semantics, indicating high levels of consistency and structural balance across languages. The research results show that this automated auditing method effectively achieves accurate assessment of the consistency of textile fiber content labeling regulations between China and Europe, and provides technical support for the construction of cross-regional regulatory harmonization and intelligent review systems.