Scholarlink

<p>Oral English proficiency is crucial for effective communication; however, most existing learning systems lack adaptability, interactivity, and personalized feedback. Conventional platforms primarily employ rule-based dialogue structures, which limits their ability to dynamically respond to diverse learner needs and varying speech patterns. To address this constraint, an intelligent English oral dialogue generation and interaction system is established using machine learning (ML). The system incorporates an Invasive Weed Optimized Intelligent CatBoost (IWO-tuned CatBoost) model, which enhances decision-making in dialogue response generation, vocabulary recommendation, and feedback adaptation. A synthetic dataset including annotated learner dialogues with grammatical and phonetic error labels supports model training. Preprocessing involves Automatic Speech Recognition (ASR)-based voice-to-text conversion, noise reduction, lemmatization, and stop word elimination. Feature extraction utilizes Mel-Frequency Cepstral Coefficients (MFCCs) for phonetic representation and Natural Language Processing (NLP)-based syntactic parsing through NLTK and spaCy for textual structure analysis. The IWO algorithm optimizes CatBoost hyperparameters to improve classification accuracy and system adaptability across diverse learner profiles. Model training is conducted in a simulated dialogue environment, enabling progressive refinement of response logic and interaction quality. The system is implemented using Python, and evaluation results indicate high performance across multiple metrics, including a BLEU score (0.82), ROUGE-L (0.79), METEOR (0.76), Engagement Score (0.84), and User Satisfaction Index (0.83), alongside a WER (0.058) and a SER (0.094). This approach demonstrates a robust and scalable framework for delivering personalized, interactive, and efficient oral English learning experiences.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Design of an english oral dialogue generation and interaction system assisted by machine learning

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月24日

查看全文
  • Zhihui Liu,
  • Jiangli Tian

摘要

Oral English proficiency is crucial for effective communication; however, most existing learning systems lack adaptability, interactivity, and personalized feedback. Conventional platforms primarily employ rule-based dialogue structures, which limits their ability to dynamically respond to diverse learner needs and varying speech patterns. To address this constraint, an intelligent English oral dialogue generation and interaction system is established using machine learning (ML). The system incorporates an Invasive Weed Optimized Intelligent CatBoost (IWO-tuned CatBoost) model, which enhances decision-making in dialogue response generation, vocabulary recommendation, and feedback adaptation. A synthetic dataset including annotated learner dialogues with grammatical and phonetic error labels supports model training. Preprocessing involves Automatic Speech Recognition (ASR)-based voice-to-text conversion, noise reduction, lemmatization, and stop word elimination. Feature extraction utilizes Mel-Frequency Cepstral Coefficients (MFCCs) for phonetic representation and Natural Language Processing (NLP)-based syntactic parsing through NLTK and spaCy for textual structure analysis. The IWO algorithm optimizes CatBoost hyperparameters to improve classification accuracy and system adaptability across diverse learner profiles. Model training is conducted in a simulated dialogue environment, enabling progressive refinement of response logic and interaction quality. The system is implemented using Python, and evaluation results indicate high performance across multiple metrics, including a BLEU score (0.82), ROUGE-L (0.79), METEOR (0.76), Engagement Score (0.84), and User Satisfaction Index (0.83), alongside a WER (0.058) and a SER (0.094). This approach demonstrates a robust and scalable framework for delivering personalized, interactive, and efficient oral English learning experiences.