Method and device for realizing automatic marking of Chinese subjective questions

Through language model pre-training and combined neural network coding, combined with high-dimensional word vector calculation, the problem of insufficient understanding of semantic relationships in automatic marking of subjective questions in Chinese subjective questions is solved, and the semantic relationships are accurately captured in different contexts, reducing the complexity of the model and hardware requirements improve the accuracy of answer recognition and the objectivity of scoring results.

CN119204008AInactive Publication Date: 2024-12-27INSPUR SOFTWARE TECH CO LTD
View PDF 7 Cites 1 Cited by

Patent Information

Application Number
CN202411718926.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-28
Publication Date
2024-12-27
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The prior art has limited understanding of semantic relationships in the automatic marking of subjective Chinese questions, and it is difficult to accurately process synonyms or synonyms, and the long-distance semantic text analysis capabilities are insufficient, which increases the complexity of the model and hardware configuration requirements, and is not suitable for use in small and medium-sized projects.

Method used

Using language model pre-training, data cleaning, combined neural network coding and high-dimensional word vector calculation, text features are extracted through masking operations and convolutional neural networks, and similarity ratio is calculated using cosine similarity to reduce model complexity and improve semantics Capture ability.

Benefits of technology

It realizes accurate capture of semantic relationships in different contexts, reduces model complexity and hardware requirements, and improves the recognition accuracy of subjective questions and the objectivity of scoring results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119204008A_ABST
    Figure CN119204008A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of artificial intelligence, and particularly provides a method and device for realizing automatic marking of Chinese subjective questions, and the method comprises the following steps: S1, model pre-training: carrying out the pre-training through word order prediction and mask operation; s2, data cleaning: processing the text by using an accurate mode of Jiababi word segmentation, and processing a complex text matching requirement by using complex regularization; s3, model coding: when the model codes the text, selecting a combined neural network to reduce the calculation amount; s4, calculating a similarity value, and calculating a similarity ratio by using cosine similarity according to the high-dimensional word vector; and S5, calculating the total score of the test paper. Compared with the prior art, the subjective question answer semantic recognition can be more accurate, and the scoring result is more objective.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Short text sentiment classification method and system based on feature enhancement

    CN113326374A

  • Multi-organ segmentation method and device for simultaneously extracting and fusing global and local features

    CN116258685A

  • Subjective question scoring method and system based on ALBERT model and RPA technology

    CN117540727A

  • Medical segmentation method based on CNN and Transform double-branch neural network model

    CN117726630A

  • Multi-modal fusion semantic perception method based on RGB and thermal infrared information

    CN118332503A

Cited By

  • High-efficiency receiving system for clothing fast reverse supply chain packaging auxiliary materials

    CN119624057A