Speaker: Giorgio Magri (Jean Nicod)
Date/Time: 4/18 (Wed) 5 pm
Location: 32D-831
Title: The OT error-driven ranking model of the acquisition of phonotactics: some computational results
Abstract:
Nine-month-old infants already react differently to licit vs illicit sound combinations, thus displaying knowledge of the target adult phonotactics. Children must thus rely on a remarkably efficient phonotactics learning strategy. What could it look like? According to the error-driven learning model, the learner maintains a current hypothesis of the target adult phonotactics and keeps slightly updating its current hypothesis whenever it makes a mistake on the incoming stream of data from the adult language. This learning model has been endorsed by the Optimality Theoretic (OT) acquisition literature because of its cognitive plausibility: it models the observed acquisition gradualness, as it describes a stepwise progression towards the target adult grammar; it relies on surface phonology without requiring any knowledge of morphology, that plausibly develops later than phonotactics; and it does not impose unrealistic memory requirements, as it only looks at a piece of data at the time without keeping track of previously seen data. Bridging cognitive plausibility with computational soundness, my current project defends the hypothesis that OT error-driven learning provides a proper model of the child acquisition of phonotactics. The project is articulated around five core issues. The first issue concerns convergence: does the model eventually stop making mistakes and settle on a final grammar? The second issue concerns correctness: does the final grammar entertained by the model at convergence indeed capture the target phonotactics? The third issue concerns modeling adequacy: do the learning sequences formally predicted by the model match attested child acquisition paths? The fourth issue concerns robustness and variation: how does the model behave in the presence of noise and how can it make sense of the pervasive phenomenon of child variation? The fifth issue concerns framework selection: how can the choice of the OT framework be justified from a learning theoretic perspective? This talk will provide an overview of the project, with a focus on some recent results concerning the first two issues of convergence and correctness.