Proceedings of the Annual Meeting of the Cognitive Science Society

Semantic Bootstrapping in Frames: A Computational Model of Syntactic Category AcquisitionAccording to the semantic bootstrapping hypothesis, children map certain (prototypical) semantic concepts to syntacticcategories (e.g., objects as nouns, actions as verbs), which are then used to learn other elements of language. We report acomputational model of syntactic category acquisition that combines semantic bootstrapping with the distributional learning oflanguage. The model has access to a small set of “seed” words, with labeled categories. It then iteratively constructs syntacticframes from the seeds; sufficiently frequent frames are used to categorize non-seeded words which then contribute to theconstruction of additional frames, including frames that incorporate category information. The model is online and effective.Simulation on child-directed English corpus shows that with only 100 seed words, classification precision exceeds 70%.