Descripción del proyecto
We learn new words almost on a daily basis: as adults, a new element is introduced in our vocabulary every other day. With new words, we also learn about new objects and ideas - in most cases new words are not simply additional labels to be applied to familiar objects: they connote meanings that are unknown to the speaker of a language. However, when we experience, as adults, an unfamiliar word, typically its referent is not immediately available in the same context. How then can language, by itself, constitute such a reliable instrument for the acquisition of novel meanings? What do we exploit to induce new meanings on the basis of an unfamiliar sequence of sounds or graphical elements? BraveNewWord addresses these questions in an innovative multidisciplinary perspective, combining cutting-edge proposals from computational linguistics and empirical investigation techniques from experimental psychology and cognitive neuroscience. BraveNewWord posits three main sources for lexically-driven meaning acquisition: linguistic context, word structure, form-meaning mapping. The project advances a computational framework that models these mechanisms through data-driven, psychologically plausible distributional systems trained on examples of natural language usage. The quantitative characterizations and algorithmic definitions offered by these models constitute, in turn, the basis for BraveNewWord large-scale empirical investigation, involving both behavioral (reaction times, mouse-tracking trajectories, diachronic language changes) and neuroscience data (event-related potentials, neuroimaging). With its innovative perspective and advanced computational and empirical approach, BraveNewWord will constitute a non-incremental contribution to understanding how human speakers use new lexical information as a mean for enriching their semantic system, and provide a ground-breaking perspective on the cognitive processes relating language and thought.