Automated Speech Scoring System Under The Lens: Evaluating and interpreting the linguistic cues for language proficiency
This work addresses the need for interpretable automated speech scoring to provide consistent feedback for second language learners, but it is incremental as it applies existing methods to analyze feature contributions.
The study tackled automated English proficiency scoring by using classical machine learning models to interpret linguistic cues, finding that regression-based models performed as well as or better than classification approaches, with ablation studies and feature importance analysis revealing the model learned underlying grading rubrics.
English proficiency assessments have become a necessary metric for filtering and selecting prospective candidates for both academia and industry. With the rise in demand for such assessments, it has become increasingly necessary to have the automated human-interpretable results to prevent inconsistencies and ensure meaningful feedback to the second language learners. Feature-based classical approaches have been more interpretable in understanding what the scoring model learns. Therefore, in this work, we utilize classical machine learning models to formulate a speech scoring task as both a classification and a regression problem, followed by a thorough study to interpret and study the relation between the linguistic cues and the English proficiency level of the speaker. First, we extract linguist features under five categories (fluency, pronunciation, content, grammar and vocabulary, and acoustic) and train models to grade responses. In comparison, we find that the regression-based models perform equivalent to or better than the classification approach. Second, we perform ablation studies to understand the impact of each of the feature and feature categories on the performance of proficiency grading. Further, to understand individual feature contributions, we present the importance of top features on the best performing algorithm for the grading task. Third, we make use of Partial Dependence Plots and Shapley values to explore feature importance and conclude that the best performing trained model learns the underlying rubrics used for grading the dataset used in this study.