Usage‐Based Language Learning
Nick C. Ellis, Matthew Brook O’Donnell, Ute Römer · 2015
Research on the learning of verb-argument constructions (VACs) emphasizes the importance of item-based patterns and their perceptual groundings in acquisition, with abstract schematic patterns emerging from the conspiracy of particular usage patterns and their interpretations. This chapter explores the distributional properties of three types of constructions (Verb-Locative, Verb Object Locative, and Verb Object Object ditransitive) as associations of form and function by means of a corpus analysis of verb selection preferences in 100 million words of usage and with the semantic network structure of the verbs in these VACs. Our initial analyses show that these constructions are (1) Zipfian in their verb type-token constituency in usage, (2) selective in their verb form occupancy, and (3) coherent in their semantics, with a network structure involving prototypical nodes of high betweenness centrality. Psychological theory relating to the statistical learning of categories suggests that these are factors that promote learning. These robust patterns of usage might therefore provide the Common Ground to facilitate processes of syntactic and semantic bootstrapping.