HedModTmplGen

Code for ACL 2019 long paper Ensuring Readability and Data-fidelity using Head-modifier Templates in Deep Type Description Generation based on OpenNMT-py.

1 Dependencies

• Python 3.5+

• PyTorch 1.0

pip install -r requirements.txt

2 Template acquicision

1) StanfordCoreNLP

Download stanford corenlp from here, and place them in the corenlp/ folder.

The start CoreNLP by running the following commands within the folder,

java -mx4g -cp "*" edu.stanford.nlp.pipeline.StanfordCoreNLPServer -port 9000

2) Run a demo

Demo can be found in raw/annotate_desc.py

python annotate_desc.py

3 Quick start

1) Dataset preparation

Quick version

Download the processed dataset from here, and place them in the data/dataNK/ folders.

More troublesome version

Download the raw infobox data from here, place them in the raw/infobox/dataNK/, and run the following to convert raw infobox to sentences.

python build_dataset.py --config config/demo-prep.yml

It may take a while. :).

2) Preprocess

Preprocess dataset into *.pt.

python preprocess.py --config config/demo-prep.yml

3) Train

python train.py --config config/demo-train.yml

4) Test

python test.py --config config/demo-test.yml

5) Evaluate

a) BLEU, ROUGE, METEOR, CIDEr

Simply add -report_bleu in test commands, or run

python evaluation.py $result_file$ $golden_file$

b) ModCopy, HedAcc

Start Stanford CoreNLP, then in the folder eval/, run

python modcopy.py -src somewhere/src-test.txt -res result_file
python hedacc.py -src somewhere/src-test.txt -tgt tgt-test.txt -res result_file

4 Citation

If you find our code or paper useful to your research, please kindly cite our paper.

@inproceedings{chen-etal-2019-ensuring,
    title = "Ensuring Readability and Data-fidelity using Head-modifier Templates in Deep Type Description Generation",
    author = "Chen, Jiangjie  and
      Wang, Ao  and
      Jiang, Haiyun  and
      Feng, Suo  and
      Li, Chenguang  and
      Xiao, Yanghua",
    booktitle = "Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics",
    month = jul,
    year = "2019",
    address = "Florence, Italy",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/P19-1196",
    doi = "10.18653/v1/P19-1196",
    pages = "2036--2046",
}