Wide-coverage probabilistic sentence processing

This paper describes a fully implemented, broad-coverage model of human syn tactic processing. The model uses probabilistic parsing techniques, which c ombine phrase structure. lexical category, and limited subcategory probabil ities with an incremental, left-to-right "pruning" mechanism based on casca ded Markov models, The parameters of the system are established through a u niform training algorithm, which determines maximum-likelihood estimates fr om a parsed corpus. The probabilistic parsing mechanism enables the system to achieve good accuracy on typical, "garden-variety " language (i.e., when tested on corpora). Furthermore, the incremental probabilistic ranking of the pre ferred analyses during parsing also naturally explains observed hum an behavior for a range of garden-path structures. We do not make strong ps ychological claims about the specific probabilistic mechanism discussed her e, which is limited by a number of practical considerations. Rather, we arg ue incremental probabilistic parsing models are, in general, extremely well suited to explaining this dual nature-generally good and occasionally path ological-of human linguistic performance.