Multi-Feature Intelligent Oral English Error Correction Based on Few-Shot Learning Technology.

Area of Science:

• Computational Linguistics
• Educational Technology
• Speech Processing

Background:

• Few-shot learning advances computer-aided language teaching (CALT) systems.
• CALT systems are increasingly used in education and assessment.
• Oral English proficiency requires accurate pronunciation evaluation.

Purpose of the Study:

• To develop a multifeature fusion-based evaluation method for oral English pronunciation.
• To improve the accuracy and usefulness of automated pronunciation scoring and feedback.
• To assist learners in enhancing their oral English pronunciation skills.

Main Methods:

• Proposed an improved scoring method based on Hidden Markov Model (HMM) a posteriori probability.
• Introduced the average pronunciation level from a corpus as an additional scoring basis, moving beyond a single reference model.
• Developed an expert opinion database for common pronunciation errors and created an artificial scoring system.

Main Results:

• The new method reduces score limitations due to individual pronunciation differences.
• The system's misjudgment rate for pronunciation errors is lowered.
• Error correction information provided to learners is more useful.
• Trials demonstrated improved scoring performance and user pronunciation levels.

Conclusions:

• The proposed method effectively evaluates specific grammar and pronunciation in oral English.
• Integrating average pronunciation levels enhances automated scoring accuracy and feedback utility.
• The system aids learners in improving their spoken English proficiency through targeted feedback.