WANG Weiwei Guangdong University of Foreign Studies

ZHANG Yuqi The Institute of Service-Oriented Manufacturing (Hangzhou) Ltd.

WANG Ke Tongyi Lab

　　Abstract:

　　This study examines the reliability of the Pretrained-model based Automatic Scoring for Interpreting (PASI) developed on the basis of the China 􀆳s Standards of English Language Ability (2024) (CSE) across different interpreting test scenarios from the dimensions of explanation, assessment and generalization, aiming to investigate the scoring quality of the PASI. The results indicate that PASI demonstrates high consistency and correlation with human scoring in homogeneous groups and simultaneous interpreting tasks, while its scoring quality still needs improvement in heterogeneous groups and consecutive interpreting tasks. It is also found that the accuracy of speech recognition significantly impacts the precision and reliability of automatic scoring. These findings suggest that automatic scoring technology has broad application prospects in the field of language testing. However, further optimization of algorithms and feature extraction models is needed to enhance scoring stability. Future research should leverage automatic scoring technology to promote the development of formative assessment in interpreting education and construct a comprehensive evaluation system that integrates formative and summative assessments.

　　Key words:

　　China's Standards of English Language Ability; Pretrained-model based Automatic Scoring for Interpreting (PASI); interpreting scoring quality; automatic scoring

　　Source:

　　Modern Foreign Languages (Bimonthly), July 2025, Vol.48 No.4