October 31, 2025 to November 2, 2025
National Olympics Memorial Youth Center / 国立オリンピック記念青少年総合センター
Asia/Tokyo timezone

Grading Lily's Students: CEFR-Aligned Automated Scoring for Video Call Conversations

Nov 2, 2025, 12:35 PM
30m
National Olympics Memorial Youth Center / 国立オリンピック記念青少年総合センター

National Olympics Memorial Youth Center / 国立オリンピック記念青少年総合センター

3-1 Yoyogikamizonocho, Shibuya, Tokyo 151-0052 / 〒151-0052 東京都渋谷区代々木神園町3-1

Speaker

Anthony Verardi (Duolingo, Inc.)

Description

Sponsored by Duolingo
Conversational AI offers new ways to practice speaking, yet reliable assessment of learner progress in real-time dialogue remains a challenge. To address this, we developed an automated grader for Video Call with Lily, Duolingo’s conversation partner. Four expert raters scored over 800 learner responses using a CEFR-aligned rubric, achieving high agreement (r=.95). Human–machine scores closely aligned (r=.85), validating the system’s reliability and establishing a foundation for scalable, data-driven feedback in self-directed language learning.

Summary

Conversational AI offers new ways to practice speaking, yet reliable assessment of learner progress in real-time dialogue remains a challenge. To address this, we developed an automated grader for Video Call with Lily, Duolingo’s conversation partner. Four expert raters scored over 800 learner responses using a CEFR-aligned rubric, achieving high agreement (r=.95). Human–machine scores closely aligned (r=.85), validating the system’s reliability and establishing a foundation for scalable, data-driven feedback in self-directed language learning.

Teaching Context General

Author

Anthony Verardi (Duolingo, Inc.)

Presentation materials

There are no materials yet.