Abstract: While syntactic dependency annotations concentrate on the surface or functional structure of a sentence, semantic depen- dency annotations aim to capture between- word relationships that are more closely related to the meaning of a sentence, using graph-structured representations. We ex- tend the LSTM-based syntactic parser of Dozat and Manning (2017) to train on and generate these graph structures. The re- sulting system on its own achieves state-of-the-art performance, beating the pre- vious, substantially more complex state- of-the-art system by 0.6% labeled F1. Adding linguistically richer input repre- sentations pushes the margin even higher, allowing us to beat it by 1.9%labeled F1.
Authors: Timothy Dozat, Christopher D. Manning (Stanford University)