Parsing Chinese by optimal contributive paths

This paper proposed an optimal contributive path, based on the probabilistic Earley algorithm. We argue that this path contributes optimally to the grammaticality of the input sentence, and hence is more capable of parsing structurally ambiguous sequences. An exponentially weighted combination of the parsing results from the optimal contributive and the Viterbi path achieved promising performances on the parsing task of CIPS ParsEval-2009, while costing only a little additional computation time since both the optimal contributive and Viterbi paths could be established simultaneously.