A Chinese character correction method and system based on attribution network and BERT

Through the attribution network and BERT method, the attribution information of characters is obtained using gradient calculation and self-attention scores, the problem of insufficient detection capabilities of Chinese character error correction tools in the prior art is solved, and higher error correction accuracy and recall rate are achieved, with stronger flexibility and versatility.

CN115809655BActive Publication Date: 2025-08-12EAST CHINA NORMAL UNIV
0 Cites 0 Cited by

Patent Information

Application Number
CN202111073538.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-09-14
Publication Date
2025-08-12
Estimated Expiration
2041-09-14

AI Technical Summary

Technical Problem

When detecting and correcting wrong characters, existing Chinese character error correction tools have problems such as insufficient detection capabilities, dependence on obfuscating set quality, complex construction, and insufficient flexibility. In particular, BERT-based models lack detection and recall capabilities in Chinese character correction.

Method used

The attribution network and BERT method is adopted to predict whether there are errors in the sentence through a binary classification model, and the attribution information of characters is obtained by using gradient calculation, and the self-attention score calculation of BERT is integrated to enhance attention to wrong characters, abandon confusion sets, and improve detection and correction capabilities.

Benefits of technology

It significantly improves the accuracy and recall of Chinese character error correction, has stronger flexibility and migration capabilities, and can effectively correct wrong characters in Chinese sentences without relying on confusion sets.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The present invention discloses a Chinese character correction method based on an attribution network and BERT. The method comprises the following steps: collecting a Chinese text dataset containing erroneous characters, and dividing the dataset into a training set and a test set for model training and testing, or using the complete dataset for training; training a BERT-based binary classification model used as an attribution network; using the trained binary classification model to perform feedforward calculations on sentences in the training set, with the output denoted as F(X); using the output F(X) to calculate the gradient of X to obtain error attribution information; setting a threshold filter to process the error attribution information obtained above; integrating the filtered error attribution information into a second BERT model for character correction training; and predicting the target character of the erroneous character. The present invention also discloses a system for implementing the above method. The method of the present invention can significantly improve the precision and recall rates in the field of Chinese text error correction, while having strong versatility and scalability.
Need to check novelty before this filing date? Find Prior Art