Multi-Grained Named Entity Recognition

This repository implements a Multi-Grained Named Entity Recognition model proposed in this paper with Tensorflow:

Congying Xia, Chenwei Zhang, Tao Yang, Yaliang Li, Nan Du, Xian Wu, Wei Fan, Fenglong Ma, Philip Yu. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics (ACL), 2019.

https://arxiv.org/abs/1906.08449

Requirements

Python

Tensorflow 1.13.1

AllenNLP

Numpy

Sklearn

Usage

Step 0.0, Prepare Data

Get Datasets.

The dataset content is subject to copyright issues. Some useful pointers:

(ACE 2004) https://catalog.ldc.upenn.edu/LDC2005T09

(ACE 2005) https://catalog.ldc.upenn.edu/LDC2006T06

(CoNLL 2003) https://cogcomp.seas.upenn.edu/page/resource_view/81

Preprocess the datasets into required data formats.

Two types of format are needed. The first example can be found at data/ace2004/train.txt. The second examples can be found at data/ace2004/elmo/sentences/train_sentences.

Please generate formated files for the whole datasets and replace the example files with your generated files, including:

data/ace2004/train.txt

data/ace2004/test.txt

data/ace2004/dev.txt

data/ace2004/elmo/sentences/train_sentences

data/ace2004/elmo/sentences/test_sentences

data/ace2004/elmo/sentences/dev_sentences

Word embeddings:

Download glove.6B.zip from https://nlp.stanford.edu/projects/glove/.

Unzip it and put glove.6B.300d.txt under the directory of data/glove.6B/

Elmo embeddings:

Run the script to download Elmo embeddings.

cd data/ace2004/elmo
sh run.sh

Three files should be generated:

elmo_dev.hdf5

elmo_test.hdf5

elmo_train.hdf5

Step 1.0, Run Detector

Step 1.1, Build data for the detector

cd detector
python build_data.py

It should generate files including

data/glove.6B.300d.trimmed.npz

data/chars.txt

data/tags.txt

data/words.txt

Step 1.2, train detector

python train.py

Step 1.3, dump shared features

python dump.py

Dumped Features will be saved in data/saved_roi/

If you'd like to evaluate the performance of the Detector, please run the evaluate.py

python evaluate.py

Step 2.0, Run Classifier

Step 2.1, train classifier

python train.py

Step 2.2, evaluate model performance

python evaluate.py

Reference

If you find our code useful, please cite our paper.

@article{xia2019multi,
  title={Multi-Grained Named Entity Recognition},
  author={Xia, Congying and Zhang, Chenwei and Yang, Tao and Li, Yaliang and Du, Nan and Wu, Xian and Fan, Wei and Ma, Fenglong and Yu, Philip},
  journal={arXiv preprint arXiv:1906.08449},
  year={2019}
}

XuanLin1991/Multi-Grained-NER