Automated Scoring of Extended Spontaneous Speech

In this chapter we provide an overview of SpeechRater ®, a state-of-the-art system for AS and feedback for extended spontaneous speech that is used operationally in a practice context by language learners preparing for a high-stakes international language test. In addition to a detailed description of the components of this automated speech scoring system, we provide the rationales and motivations for key choices in its design and system architecture. Furthermore, we place SpeechRater in the context of automated speech scoring in general, discuss how differences in task types and response characteristics, and use scenarios necessitate different choices with respect to the design and implementation of different system components. We conclude the chapter with a critical discussion of the current state-of-the-art of spontaneous speech scoring, its remaining challenges, and an outlook into future research as well as operational practice in this area.