Location: University at Buffalo

In K–12 education, Automatic Speech Recognition (ASR) is increasingly being applied to support learning, engagement, and accessibility. However, developing ASR systems for children continues to pose as a monumental challenge, especially considering the stringent requirement resulting from various use cases in children’s education settings. Moreover, representative speech datasets from these populations are particularly scarce, while data collection is further complicated by ethical considerations such as privacy and informed consent.

This invitation only workshop provides a timely forum for interested stakeholders to come together to:

• Assess the state-of-the-art in children’s ASR
• Find a common ground in our collective understanding of the needs and challenges of children’s ASR in education
• Share lessons and best practices
• Standardize evaluation and benchmarking methodologies
• Charter new (or call-for-attention to important) research directions
• Build a coalition of shared minds to collectively address those challenges for the better development of children’s ASR for education settings

We invite participants from multidisciplinary communities, to join this discussion including educators, policymakers, ed-tech entrepreneurs, speech researchers, AI-ML experts, data scientists, linguists. Given your expertise in this area, we invite you to contribute to topics relevant to this workshop, including, but not limited to:

• Diversity and quality of existing public children’s speech data sets
• Ontology development for children’s speech datasets
• Equipment and systems for children’s speech data collection
• Evaluation methodologies for existing ASR systems on children’s speech
• Needs for multi-level ASR systems for children’s speech
• Educational applications spanning age groups, subject areas, language learning contexts, and delivery formats (e.g., in-person vs. virtual)
• Initiatives and requirements for making proprietary children’s speech data publicly accessible
• Current benchmark creation and dataset design efforts advancing ASR for children’s speech
• Privacy-preserving ASR and edge ASR

This workshop presents an exceptional opportunity for an interdisciplinary, cross-sector community to collectively shape the future of children’s ASR technologies in education. Together, we aim to develop novel AI solutions that can help improve learning outcomes for all students. Your participation and contribution will enrich these discussions and amplify our collaborative investment and efforts in building a better learning environment for our children.

For inquiries about this workshop, please email Dr. Jinjun Xiong (jinjun@buffalo.edu) directly.

Please sign-up to participate in this inaugural workshop