Release of bulk commodity digital supply chain system based on Feizhu and Wenxin big model digital chain technology

Recently, the Digital Chain Technology team launched an innovative digital solution for non-standard documents based on propeller technology, aiming to break down information barriers in the commodity industry.Previously, logistics bills, contracts, and other transaction information in the commodity supply chain often existed in physical form, requiring manual photography, archiving, and information entry, resulting in issues of opacity and non-standard

Recently, the Digital Chain Technology team launched an innovative digital solution for non-standard documents based on propeller technology, aiming to break down information barriers in the commodity industry.

Previously, logistics bills, contracts, and other transaction information in the commodity supply chain often existed in physical form, requiring manual photography, archiving, and information entry, resulting in issues of opacity and non-standard. In addition, there are differences in document standards between different enterprises, resulting in a large number of offline non-standard documents, making it difficult for manual entry and traceability.

It is reported that the Digital Chain Technology team has recently launched a non-standard document digitization solution based on flying propeller technology, which mainly solves the problem of non-standard documents in the bulk commodity supply chain. Through text recognition and information extraction technology, physical documents are efficiently and accurately digitized.

Specifically, the digital chain technology team chose to use PaddleOCR, a development kit for propeller character recognition, PaddleNLP, a natural language processing model library, and UIE-X, a general information extraction model, to carry out its own research project. The team conducts research on business document types, sets key information extraction rules, and conducts model analysis and training. This solution continuously optimizes model performance through iterative data annotation, model evaluation, and business feedback.

It is reported that the application of propeller technology has improved the efficiency of non-standard document recognition, which is 20 times higher than traditional methods. The deployment process of this solution is simple, supports multiple server deployments, and uses semi precision inference to accelerate inference speed. The team utilized open-source models and tools during their self-development process, reducing the difficulty of development and deployment. (Jiang Miao)