NeurboParser

This software implements the semantic dependency parsers described in [1]: './NeurboParser' corresponds to the basic model, and './MeurboParser' to the multitask models.

Required software

The following software are needed to build the parser:

• A C++ compiler supporting the [C++11 language standard] (https://en.wikipedia.org/wiki/C%2B%2B11)
• Boost libraries (tested with version 1.61.0)
• Eigen (newer versions strongly recommended. tested with development version, changeset 9647:9464b6f3131c)
• CMake (tested with version 3.6.2)
• DyNet

Checking out the project for the first time

git clone https://github.com/Noahs-ARK/NeurboParser.git cd NeurboParser

The first time you clone the repository, you need to initialize the dynet/ submodule.

git submodule update --init

To fetch all the other required libs (except boost):

./install_deps.sh

To build the parser

mkdir -p NeurboParser/build
cd NeurboParser/build
cmake ..; make -j4
cd ../..

Running/training the parser

Follow the instructions under './semeval2015_data' to prepare the data. You can use the scripts in './NeurboParser' to train/evaluate the parser.

To replicate the results

You will first need to put 100-dimensional pretrained GloVe embedding [3] under 'embedding/'. Default hyperparameters in the scripts are used in [1].

The current version of DyNet uses some different strategies to deal with numerical issues than the older version we used in [1]. Based on our experience, we expect the current parser to have slightly better evaluation numbers on benchmark datasets than those described in [1].

We are still working on adapting the multitask models to use the new version of DyNet.

References

[1] Hao Peng, Sam Thomson, and Noah A. Smith. 2017. Deep Multitask Learning for Semantic Dependency Parsing In Proceedings of the Annual Meeting of the Association for Computational Linguistics (ACL).

[2] André F. T. Martins, Miguel B. Almeida, Noah A. Smith. 2013. Turning on the Turbo: Fast Third-Order Non-Projective Turbo Parsers. In Proceedings of the Annual Meeting of the Association for Computational Linguistics (ACL).

[3] Jeffrey Pennington, Richard Socher, and Christopher D. Manning. 2014. GloVe: Global Vectors for Word Representation. In Proceedings of the Empirical Methods in Natural Language Processing (EMNLP).

[4] André F. T. Martins and Mariana S. C. Almeida. 2014. Priberam: A Turbo Semantic Parser with Second Order Features. In Proceedubgs of the International Workshop on Semantic Evaluation (SemEval), task 8: Broad-Coverage Semantic Dependency Parsing.

[5] Mariana S. C. Almeida and André F. T. Martins. 2015. Lisbon: Evaluating TurboSemanticParser on Multiple Languages and Out-of-Domain Data. In Proceedings of International Workshop on Semantic Evaluation (SemEval'15), task 18: Broad Coverage Semantic Dependency Parsing.