Accounting document classification method and system based on artificial intelligence

By combining the feature fusion of text and image modalities, using the self-attention mechanism and tensor fusion algorithm to generate trusted labels, the problem of insufficient accuracy of single modals in accounting voucher classification is solved, and more efficient multi-label classification is achieved.

CN120372010AActive Publication Date: 2025-07-25JILIN TECH COLLEGE OF ELECTRONICS INFORMATION
View PDF 6 Cites 1 Cited by

Patent Information

Application Number
CN202510400439.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-04-01
Publication Date
2025-07-25
Estimated Expiration
2045-04-01

AI Technical Summary

Technical Problem

In the prior art, the single-modal classification method of accounting vouchers is prone to degradation of classification accuracy due to image quality problems or unclear text information, and it is difficult to effectively combine text modes and image modes for multi-label classification.

Method used

By obtaining the tag data set of accounting vouchers, extracting text information and voucher images, semantic word segmentation and semantic association matching, combining self-attention mechanism and tensor fusion algorithm, fusing images and text features to generate trusted tags for multi-label classification.

Benefits of technology

It improves the accuracy and semantic understanding of accounting voucher classification, and makes up for the information blind spots of single modal classification. Especially when the image quality is poor, text information can effectively supplement the missing parts of the image and enhance the performance of multi-label classification.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120372010A_ABST
    Figure CN120372010A_ABST
Patent Text Reader

Abstract

The invention provides an accounting document classification method and system based on artificial intelligence, and the method comprises the steps: determining a label dependency relationship between text information and a label data set according to the part-of-speech of effective lexical elements in each text keyword; determining text label features in the target accounting document through the label dependency relationship and the semantic loss of each text keyword; performing semantic association matching on the semantic features of the voucher image and each data tag in the tag data set to obtain a semantic inline relationship between the voucher image and the tag data set, and determining image tag features of semantic lexical elements in the target accounting voucher based on the semantic inline relationship; and performing modal fusion on the image tag features and the text tag features, generating a credible tag of the target accounting document based on the tag features after modal fusion, and performing multi-tag classification on the target accounting document by using the credible tag. Based on the scheme, multi-label classification of accounting documents can be realized in combination with feature fusion of the text modality and the image modality.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Automatic comment generation method based on large-scale language model and related device

    CN116882414A

  • Text prediction method and device, equipment and storage medium

    CN116975199A

  • Data information label processing method of large language model

    CN117453921A

  • Text classification method and device based on semantics, computer equipment and storage medium

    CN117874234A

  • Conversation intention recognition method, apparatus, computer device, and storage medium

    WO2021042904A1

Cited By

  • Self-adaptive agent service platform based on unstructured data intelligent analysis

    CN121170345A