Field: Language Learning

Research Question:
Do words acquired later in first-language development (L1) show higher error rates in second language (L2) practice on Duolingo?

Duolingo SLAM Dataset:
• Large corpus of data from over 6,000 Duolingo users, collected during their first 30 days of learning a language
• Released in 2018 as part of the Duolingo Shared Task on Second Language Acquisition Modeling (SLAM).
• Publicly available via Harvard Dataverse and linked from Duolingo Research.
• Freely usable for research/educational purposes (requires agreeing to terms of use).

WordBank Dataset:
• Open database of children’s vocabulary growth.
• Publicly available at wordbank.stanford.edu.
• Aggregates data from the MacArthur–Bates Communicative Development Inventories (CDIs).
• Open access under a permissive license (for research/educational use).