Multi-modal test question knowledge point labeling method, system and equipment and storage medium

The multi-modal knowledge point annotation method addresses the inefficiencies of existing annotation methods by integrating self-supervised dual-stream feature extraction and cross-modal alignment networks to generate comprehensive feature vectors, enhancing annotation accuracy and supporting personalized learning.

CN119720945AActive Publication Date: 2025-03-28NORTHWESTERN POLYTECHNICAL UNIV
2 Cites 2 Cited by

Patent Information

Application Number
CN202411843810.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-12-14
Publication Date
2025-03-28
Estimated Expiration
2044-12-14

AI Technical Summary

Technical Problem

Current knowledge point annotation methods in intelligent education rely heavily on human expertise and are inefficient, struggling to effectively integrate and align textual and visual information from diverse educational resources, limiting their applicability and accuracy across various subjects.

Method used

A multi-modal knowledge point annotation method that employs a self-supervised dual-stream feature extraction network, followed by a two-stage cross-modal alignment network and a difference-aware fusion network to integrate textual and visual features, generating a comprehensive feature vector for accurate annotation.

Benefits of technology

This approach significantly enhances the efficiency and accuracy of knowledge point annotation by effectively aligning and fusing textual and visual modalities, enabling automated annotation and supporting personalized learning, adaptive question recommendation, and improved teaching services.

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

Abstract

The invention discloses a multi-modal test question knowledge point labeling method, system and device based on image-text fusion and a storage medium, and belongs to the technical field of natural language processing. The method comprises the following steps: performing feature extraction on test question data through a self-supervised auxiliary double-flow test question feature extraction network to obtain a text feature vector and an image feature vector of a test question; inputting the text feature vector and the image feature vector into a two-stage cross-modal alignment network for alignment to obtain an aligned text feature vector and an aligned image feature vector; inputting the aligned text feature vector and the aligned image feature vector into a difference perception fusion network to obtain a fusion vector; performing weighted calculation on the fusion vector, the text feature vector and the image feature vector to obtain a test question feature vector; and inputting the test question feature vectors into a knowledge point labeling network to obtain knowledge point labels, and labeling the questions. According to the invention, the accuracy and efficiency of knowledge point marking can be improved, and more accurate and personalized support can be provided for teaching services.
Need to check novelty before this filing date? Find Prior Art