禎洋 Funded by Online Language Learning AI Assistant that Grows with People (JPNP20006), New Energy and Industrial Technology Development Organization (NEDO), 2020-2024. XR Communication Infrastructure for Realizing High-Immersion Interaction Experiences with Conversational AI Agents (JPJ012368C06301), the National Institute of Information and Communications Technology (NICT), 2022-2024.
et al. 2024] 英会話におけるスピーキング能力判定と会話練習サービスを教育団体やビジネス業界へ展開 7 [Saeki et al. 2024] Mao Saeki, Hiroaki Takatsu, Fuma Kurata, Shungo Suzuki, Masaki Eguchi, Ryuki Matsuura, Kotaro Takizawa, Sadahiro Yoshikawa, and Yoichi Matsuyama. "InteLLA: Intelligent Language Learning Assistant for Assessing Language Proficiency through Interviews and Roleplays." In Proceedings of the 25th Annual Meeting of the Special Interest Group on Discourse and Dialogue, pp. 385-399. 2024 Best Paper Award.
can use. Accuracy How correctly the speaker uses language structures (grammar, vocabulary). Coherence How logically ideas are organized and linked in speech. Phonology Pronunciation, intonation, and how well sounds are articulated. Fluency The ability to speak smoothly without unnecessary pauses or hesitation. Interaction The ability to engage with others, responding appropriately and managing conversation. スピーキング能力判定 | CEFRの定義 Overall The combined ability to communicate effectively in spoken language. A1 A2 B1 B2 C1 C2 Beginner Mastery Phonology Coherence Fluency Accuracy Interactio n Range
Warm-up Level check Probe Level check Probe Cool down Level Up 対話破綻 対話破綻 Topics CEFR Level ACTFL OPI [Liskin-Gasparro 2003] のプロトコルを基に、システムはユーザーの言語習熟度に応じてトピッ クの複雑さをリアルタイムに調整しながら対話を進める Topic 最近の休日の出来事 について語る、または 身近なテーマについ て意見を述べる [Liskin-Gasparro 2003] J Liskin‐Gasparro, Judith E. "The ACTFL proficiency guidelines and the oral proficiency interview: A brief history and analysis of their survival." Foreign Language Annals 36, no. 4 (2003): 483-490. Level Up
少ないリソースでの動作 システムはさまざまな環境で動作し、低スペックのクライアントデ バイスからアクセスできる必要がある 能力判定手法 | システムの必要条件 [Saeki et al. 2024] Mao Saeki, Hiroaki Takatsu, Fuma Kurata, Shungo Suzuki, Masaki Eguchi, Ryuki Matsuura, Kotaro Takizawa, Sadahiro Yoshikawa, and Yoichi Matsuyama. "InteLLA: Intelligent Language Learning Assistant for Assessing Language Proficiency through Interviews and Roleplays." In Proceedings of the 25th Annual Meeting of the Special Interest Group on Discourse and Dialogue, pp. 385-399. 2024 Best Paper Award.
2024] モジュラー設計 : 複数モジュールを同時に動作させるアーキテクチャを採用し、各モジュールが連携することで高速 なリアルタイム処理を実現 サーバサイドレンダリング : さまざまな環境のクライアントデバイスでの動作を保証するため、バーチャルエージェ ントはサーバレンダリングし、WebRTCによりユーザーとの双方向通信を実現 [Saeki et al. 2024] Mao Saeki, Hiroaki Takatsu, Fuma Kurata, Shungo Suzuki, Masaki Eguchi, Ryuki Matsuura, Kotaro Takizawa, Sadahiro Yoshikawa, and Yoichi Matsuyama. "InteLLA: Intelligent Language Learning Assistant for Assessing Language Proficiency through Interviews and Roleplays." In Proceedings of the 25th Annual Meeting of the Special Interest Group on Discourse and Dialogue, pp. 385-399. 2024 Best Paper Award.
DialOpsの整理 | MLOpsとの共通点 [Sculley et al. 2015] D. Sculley, Gary Holt, Daniel Golovin, Eugene Davydov, Todd Phillips, Dietmar Ebner, Vinay Chaudhary, Michael Young, Jean-Francois Crespo, and Dan Dennison. Hidden technical debt in machine learning systems. In Proceedings of the 28th International Conference on Neural Information Processing Systems - Volume 2, NIPS’15, page 2503–2511, Cambridge, MA, USA, 2015.
al. 2023] Fuma Kurata, Mao Saeki, Shinya Fujie and Yoichi Matsuyama, Multimodal turn-taking model using visual cues for end-of-utterance prediction in spoken dialogue systems, Proc. Interspeech 2023, pp.2658-2662, 2023. 会話中の騒音への対処は大きな課題のひとつであり、その問題へ対処したモデル