雅思口语AI评分模型迭代——雅思口语四项评分维度分数与口语总分关系的研究

Iterative Development of AI Scoring Models for IELTS Speaking—A Study on the Relationship between Sub-Scores of Four Rating Demensions and Overall Speaking Score

摘要: 雅思口语AI模考及评分系统正在教学中扮演愈发重要的角色，其有效提高了教师批改作业和对学生进行个性化反馈的效率，因此提升AI模考系统的评分准确性是一个重要的研究方向。本文通过对新东方八个分中心雅思教师的实际考试口语总分与对应的四项评分维度分数进行数据建模，得出了通过4项维度分数计算口语总分的加权模型公式。本文也呈现了部分统计数据，为后续口语模考平台迭代与学生针对性提分练习提供了支持。

Abstract: AI-powered IELTS speaking mock test and scoring system is playing an increasingly important role in language teaching and has effectively improved the educators’ efficiency in grading assignments and giving personalized feedback to students. Consequently, enhancing the accuracy of the mock test system is a vital focus of further research. This paper analyzes the real test scores from IELTS teachers in eight New Oriental Sub-centers and establishes a weighted formula for calculating the overall IELTS Speaking band from its four score breakdowns. Other than that, there are statistical findings presented to support the potential improvements of the AI-powered mock test platform and focused practice strategies for students.