Intelligent text analysis for effective evaluation of english Language teaching based on deep learning

Area of Science:

• Natural Language Processing
• Artificial Intelligence in Education

Background:

• The demand for English language teaching necessitates efficient and accurate student writing evaluation.
• Traditional automated essay scoring methods struggle with cross-prompt adaptability and deep semantic understanding.

Purpose of the Study:

• To introduce a Hybrid Feature-based Cross-Prompt Automated Essay Scoring (HFC-AES) model for intelligent text analysis.
• To enhance the robustness and semantic modeling capabilities of automated essay scoring systems.
• To provide reliable and efficient evaluation tools for English language teaching.

Main Methods:

• Developed a Hybrid Feature-based Cross-Prompt Automated Essay Scoring (HFC-AES) model using deep learning.
• Incorporated text structure features and attention mechanisms into deep neural networks (DNNs).
• Employed adversarial training for optimized feature extraction and cross-prompt adaptability, with topic-independent and topic-specific scoring stages.

Main Results:

• The HFC-AES model achieved a strong cross-prompt scoring performance with an average Quadratic Weighted Kappa (QWK) of 0.856.
• HFC-AES outperformed existing Transformer-based scoring models in robustness and semantic modeling.
• Ablation studies confirmed the importance of text structure features and attention mechanisms, especially for argumentative writing.

Conclusions:

• The HFC-AES model offers effective technical support for automated essay grading.
• The model contributes to more reliable and efficient evaluation in English language teaching.
• The study highlights the potential of hybrid feature approaches and deep learning for advanced automated essay scoring.