大语言模型背景下提示词工程赋能英语口语学习研究

Empowering Spoken English Learning with Prompt Engineering against the Background of Large Language Models

摘要: 本研究探讨了运用提示词工程和大语言模型来适应各种英语口语教育场景，研究了为模拟雅思口语考试、K-12口语考试和学前英语教学等场景而定制的提示词的设计和应用。本研究使用了定量指标，如BLEU和ROUGE分数，以及流利性、相关性、完整性、多样性和连贯性等指标，评估生成的回复的质量。此外，本研究还运用了综合评分公式来评估回复的质量，并说明了每个指标的重要性。本研究强调了提示词工程在为英语教育提供适应性解决方案方面的潜力。它还强调了需要进一步改进大语言模型的能力，以提高其在这些场景中的性能。本研究为提示词工程和大语言模型的应用提供了宝贵的见解，为英语口语教育及相关领域的发展做出了贡献。

Abstract: This research explores the utilization of prompt engineering and large language models to adapt to various English oral education scenarios. It investigates the design and application of prompts tai-lored to simulate scenarios such as IELTS speaking tests, K-12 oral exams, and preschool English teaching. The study assesses the quality of generated responses using quantitative metrics, including BLEU and ROUGE scores, fluency, relevance, completeness, diversity, and coherence. Furthermore, it introduces a comprehensive scoring formula to evaluate response quality, accounting for the significance of each metric. Despite some limitations, the research highlights the potential of prompt engineering in providing adaptable solutions for English language education. It also under-scores the need for further improvements in LLMs’ capabilities to enhance their performance in these scenarios. This study contributes to the advancement of English oral education and related fields by providing valuable insights into the application of prompt engineering and large language models.