Customs import and export commodity specification declaration review model that integrating VAE and BERT
Peng Zhang, Chao Che, Qiang Zhang · 2022 IEEE International Conference on Advances in Electrical Engineering and Computer Applications (AEECA) · 2022
Standardized declaration review refers to judging whether the declaration information of import and export commodities filled in by enterprises meets the customs reporting standards in accordance with the standardized declaration catalogue formulated by the general administration of customs. Commodity declaration information is a short text composed of independent elements, and there are problems such as scattered keywords, semantic discontinuity, and elements that are not easy to identify and split. In view of these problems, this paper proposes a model that integrates variational encoders and BERT to review the product declaration text. The model reconstructs the text through a variational encoder, combines the word frequency to capture keywords under different elements of the text, so that the model learns the unique structure of the commodity text. Make full use of the customs corpus, combine and expand the BERT thesaurus, and use the tokenization method of word granularity to correctly divide the customs vocabulary, so that the extracted element information is more complete. The experimental results on the real customs commodity declaration data show that the proposed model is conducive to improving the intelligent level of standardized declaration and review, and the accuracy of the prediction results is relatively high.