Automated Scoring and Feedback for Spoken Language
Klaus Zechner, Ching‐Ni Hsieh · 2024
This chapter provides an overview of state-of-the-art automated scoring of spoken responses in English language assessments, including a detailed description of the major components necessary for such systems: automatic speech recognition, feature computation, identification of non-scoreable responses, and scoring models. Differences between the automated scoring of written language (such as essays) and spoken language are discussed, providing a context for this chapter. Additionally, the chapter lays out the background and examples related to providing feedback to language learners that go beyond a single numeric score for a response. Finally, recent developments in automated speech scoring that are not based on curated features but on end-to-end models using deep neural networks are critically reviewed.