Please join us this Thursday at 4:15 to hear a practice presentation for BU by Tim Hunter and Laurel Perkins. Details below.

Also note that the following week, when some people are away at BU, the rest of us will do a “journal club” meeting, so please come to this week’s meeting with any suggestions for articles you want to read/present, in case there’s time at the end to discuss that. If not, we will have the discussion online.

Word order acquisition from immature representations: A model for noise-tolerant learning of deterministic grammars

Children acquire their language’s canonical word order early in infancy, despite immature abilities to identify clause arguments. Prior work proposes that learning requires prosodic or semantic support. We demonstrate that it is possible for learners to learn word order only from strings of imperfectly-identified noun-phrases and verbs. Our modelled learner chooses among grammars that deterministically produce a canonical word order, while also considering probabilistic processes that introduce noise into the data. In simulations on child-directed English and French, the model successfully separated signal from noise in order to identify canonical SVO word order in both languages. We find that successful learning does not require prosodic or semantic cues to sentence structure, but benefits from determinism in the learner’s hypothesis space. This provides an alternative mechanism for regularization in learning, where tendencies to regularize emerge from the learner’s expectation that its data are a noisy realization of a deterministic underlying system.