The application discloses a document intelligent sorting
system and method based on size model combination, and relates to the technical field of document intelligent
processing. The
system comprises a
template library preparation module, a single-page intelligent classification module, a first-round sorting module, a
large model secondary sorting module, a result feedback module, a
knowledge base management module and a model fine-tuning module. First, the document template is subjected to OCR recognition, and template
directory information and keyword information are extracted to construct a
template library. Then, each document page of the document to be sorted is subjected to OCR recognition, and the preliminary
classification result is determined based on the cross-validation result of the template page similarity and the keyword coverage. For the classification grouping of the fixed page number template, the rule check is performed, and for the fixed page number template and the non-fixed page number template with repeated pages or missing pages, the
large model is combined with the unclassified grouping data, the related document template
directory and the content of each page to perform secondary sorting. Finally, the positive
knowledge base and the negative
knowledge base are constructed through the feedback of business experts, and the Embedding model is fine-tuned based on the positive and negative knowledge bases, so that the self-iteration
upgrade of the
system is realized. The application can improve the document sorting efficiency, accuracy and generalization ability, and has the
continuous optimization capability.