Text correction method and apparatus, and computer device, storage medium and program product

By using a teacher-student model architecture and a lightweight Transformer model, and combining semantic distance and contextual information to train the student model, the problem of low reliability in traditional text correction methods is solved, and more accurate and consistent text correction results are achieved.

WO2026108786A1PCT designated stage Publication Date: 2026-05-28CHINA TELECOM CLOUD TECH CO LTD
View PDF 5 Cites 0 Cited by

Patent Information

Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
CHINA TELECOM CLOUD TECH CO LTD
Filing Date
2025-11-17
Publication Date
2026-05-28

AI Technical Summary

Technical Problem

In traditional text correction methods, newly generated words may not conform to the semantics of the text, resulting in low reliability of error correction.

Method used

A teacher-student model architecture is adopted, in which the student model is trained by the teacher model and sample data. The parameters of the student model are adjusted by combining semantic distance and contextual information to improve the error correction accuracy, and a lightweight Transformer model is used to improve the generalization ability.

Benefits of technology

It improves the reliability of text correction, ensuring that the corrected text remains coherent and logical in context, and enhances the efficiency and accuracy of correction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN2025135503_28052026_PF_FP_ABST
    Figure CN2025135503_28052026_PF_FP_ABST
Patent Text Reader

Abstract

A text error correction method, comprising: acquiring target text to be corrected (S101); and further inputting said target text into a target error correction model, so as to obtain target corrected text (S102), wherein the target error correction model is obtained by training a student model on the basis of a teacher model and sample data, and the sample data comprises sample text to be corrected, sample correct text corresponding to said sample text, and context information corresponding to the sample correct text.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Chinese text error correction method and device and computer equipment

    CN115496074A

  • Training method of grammar error correction model and related device

    CN117217210A

  • Text error correction model training method and device, text error correction method and device, equipment and medium

    CN117743857A

  • Text error correction method and device, computer equipment, storage medium and program product

    CN119578409A

  • Code generation through reinforcement learning using code-quality rewards

    US20230195428A1