A corpus of ~290 hours of individual and small-group math and computer science tutoring session transcripts with middle and high-school students.
Transcripts are being de-identified, cleaned, and standardized into a turn-level schema (JSONL), segmented into turns, exchanges, and problem episodes, and annotated.
Tutor move annotations (focus, probing, telling, generic) are adapted from MathDial (Macina et al., 2023) and student engagement (passive, active, constructive, interactive) uses the ICAP framework (Chi & Wylie, 2014).
From August 2025 to November 2026, ~200 hours of computer science and ~90 hours of mathematics tutoring have been recorded.
All sessions are recorded with student and guardian consent. Data kept private due to identification risk.