Left-to-Right Dependency Parsing with Bottom-up Hierarchical Pointer Networks

This repository includes the code of the left-to-right dependency parser with Bottom-up Hierarchical Pointer Networks described in the paper Dependency Parsing with Bottom Hierarchical Pointer Networks. The implementation is based on the dependency parser by Liu et al. (2019) (https://github.com/ntunlp/ptrnet-depparser) and reuses part of its code, including data preparation and evaluating scripts.

Requirements

Python 2.7, PyTorch >=0.4.1, Gensim >= 0.12.0

Experiments

1. Update examples/run.sh with the paths for data and embeddings.
2. Run command ./examples/run.sh <log_name>.

Citation

@article{FERNANDEZGONZALEZ2023494,
title = {Dependency parsing with bottom-up Hierarchical Pointer Networks},
journal = {Information Fusion},
volume = {91},
pages = {494-503},
year = {2023},
issn = {1566-2535},
doi = {https://doi.org/10.1016/j.inffus.2022.10.023},
url = {https://www.sciencedirect.com/science/article/pii/S1566253522001993},
author = {Daniel Fernández-González and Carlos Gómez-Rodríguez},
keywords = {Natural language processing, Computational linguistics, Parsing, Dependency parsing, Neural network, Deep learning},
abstract = {Dependency parsing is a crucial step towards deep language understanding and, therefore, widely demanded by numerous Natural Language Processing applications. In particular, left-to-right and top-down transition-based algorithms that rely on Pointer Networks are among the most accurate approaches for performing dependency parsing. Additionally, it has been observed for the top-down algorithm that Pointer Networks’ sequential decoding can be improved by implementing a hierarchical variant, more adequate to model dependency structures. Considering all this, we develop a bottom-up oriented Hierarchical Pointer Network for the left-to-right parser and propose two novel transition-based alternatives: an approach that parses a sentence in right-to-left order and a variant that does so from the outside in. We empirically test the proposed neural architecture with the different algorithms on a wide variety of languages, outperforming the original approach in practically all of them and setting new state-of-the-art results on the English and Chinese Penn Treebanks for non-contextualized and BERT-based embeddings.}
}

Acknowledgments

We acknowledge the European Research Council (ERC), which has funded this research under the European Union’s Horizon 2020 research and innovation programme (FASTPARSE, grant agreement No 714150) and the Horizon Europe research and innovation programme (SALSA, grant agreement No 101100615), ERDF/MICINN-AEI (SCANNER-UDC, PID2020-113230RB-C21), Xunta de Galicia (ED431C 2020/11), and Centro de Investigación de Galicia "CITIC", funded by Xunta de Galicia and the European Union (ERDF - Galicia 2014-2020 Program), by grant ED431G 2019/01.