ABSTRACT
Since interactive speaking in a conversational setting is required for English learners to communicate effectively in both workplace and academic environments, the inclusion of tasks that elicit dialogic speech in an English speaking proficiency assessment is important for achieving more complete construct coverage and test authenticity. While computer-based speaking proficiency assessments have typically only included monologic prompt-response speaking tasks, recent improvements in Spoken Dialog System (SDS) technology have made it possible to also consider use of SDS-based, dialogic conversational tasks in next-generation speaking assessments. While SDS-based speaking tasks have the potential to improve automated assessment systems by eliciting more authentic speech and opening up ability to assess aspects of the speaking construct that are not covered by monologic speaking tasks, there are many challenges associated with introducing them into consequential assessments. In some cases, issues can be addressed by providing a clarification prompt to the user if the learner’s response does not match any of the expected semantic categories.
