A Hybrid Japanese Parser with Hand-crafted Grammar and Statistics

This paper describes a hybrid parsing method for Japanese which uses both a hand-crafted grammar and a statistical technique. The key feature of our system is that in order to estimate likelihood for a parse tree, the system uses information taken from alternative partial parse trees generated by the grammar. This utilization of alternative trees enables us to construct a new statistical model called Triplet/Quadruplet Model. We show that this model can capture a certain tendency in Japanese syntactic structures and this point contributes to improvement of parsing accuracy on a shallow level. We report that, with an underspecified HPSG-based grammar and a maximum entropy estimation, our parser achieved high accuracy: 88.6% accuracy in dependency analysis of the EDR annotated corpus, and that it outperformed other purely statistical parsing methods on the same corpus. This result suggests that proper treatment of hand-crafted grammars can contribute to parsing accuracy on a shallow level.