A new study published on arXiv investigates the roles of lexical frequency and contextual surprisal in Spanish language acquisition and adult reading. The research found that lexical frequency was a strong predictor of when children acquire individual words, while contextual surprisal, derived from language models like BETO, BERTIN, and mGPT, added little to this prediction. However, in adult reading, surprisal robustly predicted longer fixation durations, suggesting that frequency is more critical for early lexical acquisition and surprisal for moment-to-moment processing in established linguistic systems. AI
IMPACT Suggests language models' surprisal metric is more relevant for adult reading than child word acquisition.
RANK_REASON Academic paper on language acquisition and processing. [lever_c_demoted from research: ic=1 ai=0.7]
- BERTIN
- BETO
- Chilean Spanish
- Francisco Portillo López
- MECO Wave 2
- Multilingual Eye-movement Corpus
- Spanish
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →