Improving Dependency Parsing with Fined-Grained Features

In this paper, we present a simple and effective fine-grained feature generation scheme for dependency parsing. We focus on the problem of grammar representation, introducing fine-grained features by splitting various POS tags to different degrees using HowNet hierarchical semantic knowledge. To prevent the oversplitting, we adopt a threshold-constrained bottomup strategy to merge the derived subcategories. We conduct the experiments on the Penn Chinese Treebank. The results show that, with the fine-grained features, we can improve the dependency parsing accuracies by 0.52% (absolute) for the unlabeled first-order parser, and in the case of second-order parser, we can improve the dependency parsing accuracies by 0.61% (absolute).