Hybrid semi-Markov CRF

HSCRF achieves F1 score of 91.38+/-0.10 on the CoNLL 2003 NER dataset, without using any additional corpus or resource.

Installation

PyTorch

The code is based on PyTorch. You can find installation instructions here.

Dependencies

The code is written in Python 2.7 and pytorch 0.2.0. Its dependencies are summarized in the file requirements.txt. You can install these dependencies like this:

pip install -r requirements.txt

Code reference

LM-LSTM-CRF

Usage

CUDA_VISIBLE_DEVICES=0 python train.py --char_lstm --high_way

word embeddings

Glove: You can find the pre-trained word embedding here,
and place glove.6B.100d.txt in data/.

Cite

If you use the code, please cite the following paper:
“Hybrid semi-Markov CRF for Neural Sequence Labeling”
Zhi-Xiu Ye, Zhen-Hua Ling. ACL (2018)

@InProceedings{HSCRF,
  author =     "Ye, Zhixiu
        and Ling, Zhen-Hua",
  title =     "Hybrid semi-Markov CRF for Neural Sequence Labeling",
  booktitle =     "Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers)",
  year =     "2018",
  publisher =     "Association for Computational Linguistics",
  pages =     "235--240",
  location =     "Melbourne, Australia",
  url =     "http://aclweb.org/anthology/P18-2038"
}