Deep Learning Scoring Model in the Evaluation of Oral English Teaching

Area of Science:

• Artificial Intelligence
• Computational Linguistics
• Educational Technology

Background:

• Oral English recognition accuracy is crucial for language learning.
• Existing speech recognition models often lack specific oral error correction.
• Deep learning offers potential for advanced speech analysis.

Purpose of the Study:

• To improve the accuracy of oral English recognition.
• To propose novel evaluation measures for oral English proficiency.
• To develop a deep learning-based scoring model for intelligent English teaching.

Main Methods:

• Constructed an English speech recognition model using recurrent neural networks.
• Integrated oral error correction standards into the model.
• Utilized 20 English majors and 5 sets of sentence patterns for testing.

Main Results:

• Achieved an average speech recognition concordance rate of 91%.
• Specific concordance rates: 89% for words, 91% for speech, and 86% for intonation.
• The model demonstrated effectiveness as an English speech recognition evaluation system.

Conclusions:

• The deep learning scoring model significantly improves oral English recognition accuracy.
• The model provides a robust basis for evaluating intelligent English teaching.
• This research offers advancements in automated oral English assessment.