Semantic Role Labeling with Iterative Structure Refinement

This repository contains code for training semantic role labeling model described in: Semantic Role Labeling with Iterative Structure Refinement

We use English as example, other languages have similar configuration.

If you use our code, please cite our paper as follows:

@inproceedings{lyu-etal-2019-semantic,
title={Semantic Role Labeling with Iterative Structure Refinement},
author={Lyu, Chunchuan and Cohen, Shay B. and Titov, Ivan},
booktitle={"Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP)"},
month={nov}
year={2019}
}

Prerequisites:

Python 3.6
pytorch 1.0
2009 CoNLL Shared Task

Training:

Train baseline model

allennlp train exps/srl2009_base.json --include-package myallennlp --serialization-dir ../Iterative_Inference_Models/en_base

Move out model file and vocabulary file for building refiner:

cp ../Iterative_Inference_Models/en_base/model.tar.gz ../Iterative_Inference_Models/base_en_model.tar.gz cp -r ../Iterative_Inference_Models/en_base/vocabulary ../Iterative_Inference_Models/en_vocabulary

Train refiner

allennlp train exps/srl2009.json --include-package myallennlp --serialization-dir ../Iterative_Inference_Models/refine

Testing

allennlp predict ../Iterative_Inference_Models/refine/model.tar.gz [../CoNLL2009-ST-evaluation-English.txt] --batch-size 128 --cuda-device 0 --use-dataset-reader --include-package myallennlp --predictor dependency_srl

Evaluation

Official transcript: https://ufal.mff.cuni.cz/conll2009-st/scorer.html [perl] eval09.pl [OPTIONS] -g -s

perl eval09.pl -g [../CoNLL2009-ST-evaluation-English.txt] -s ../Iterative_Inference_Models/refine/CoNLL2009-ST-evaluation-English.predict -q

Contact

Contact (chunchuan.lv@gmail.com) if you have any questions!