code and dataset for AAAI 2021 paper: Enhancing Scientific Papers Summarization with Citation Graph
we recommend using the following commands to install the PYROUGE environment:
sudo apt-get install libxml-perl libxml-dom-perl
pip install git+git://github.com/bheinzerling/pyrouge
export PYROUGE_HOME_DIR=the/path/to/RELEASE-1.5.5
pyrouge_set_rouge_path $PYROUGE_HOME_DIR
chmod +x $PYROUGE_HOME_DIR/ROUGE-1.5.5.plYou can refer to https://github.com/andersjo/pyrouge/tree/master/tools/ROUGE-1.5.5 for RELEASE-1.5.5 and remember to build Wordnet 2.0 instead of 1.6 in RELEASE-1.5.5/data\
cd $PYROUGE_HOME_DIR/data/WordNet-2.0-Exceptions/
./buildExeptionDB.pl . exc WordNet-2.0.exc.db
cd ../
ln -s WordNet-2.0-Exceptions/WordNet-2.0.exc.db WordNet-2.0.exc.dbThe whole dataset and its corresponding citation relationship can be download through this link
example of our dataset:
{
"paper_id": "102498304", # unique id of this paper
"title":"Weak Galerkin finite element method for Poisson’s ...", # title of this paper
"abstract":"in this paper , the weak galerkin finite element method for second order eilliptc problems employing polygonal or ...", # human written abstract
"text":[
["The weak galerkin finite element method using triangulated meshes was proposed by .."],
["Let @inlineform1 be a partition of the domain Ω consisting of polygons in two dimensional"],
...
] # body text,
"section_names": ["Introduction", " Shape Regularity", ...] # corresponding section names to sections
"domain":"Mathematic", # class label
}
...
You can download our preprocessed dataset which can be directly loaded by dataloader.py via SSN (inductive) and SSN (transductive).
Note that we divide the dataset in two ways. The transductive division indicates that most neighbors of papers in test set are from the training set, but considering that in real cases, the test papers may from a new graph which has nothing to do with papers we used for training, thus we introduce SNN (inductive), by splitting the whole citation graph into three independent subgraphs – training, validation and test graphs.
Our preprocessed datasets are chunked to 500 words, for full document you can retrieve them from the whole dataset by paper_id
你也可以通过百度云下载我们的数据集 SSN完整数据集和引用关系 提取码 v4u8
SSN inductive 提取码 gk4j
SSN transductive 提取码 17kw
Hyperparameters in the train.py/test.py script has been set to default, we also provide the example to run our code in train.sh and test.sh.
you can train/test our model using the following command:
- training
python train_CGSum.py --visible_gpu 0 --model_dir save_models/CGSum_1hop --dataset_dir SSN/inductive --setting inductive --n_hop 1python train_CGSum.py --visible_gpu 0 --model_dir save_models/CGSum_1hop --dataset_dir SSN/transductive --setting transductive --n_hop 1- testing
python test_CGSum.py --visible_gpu 0 --model_dir save_models/CGSum_1hop --model_name CGSum_inductive_1hopNbrs.pt --setting inductive --decode_dir decode_path --result_dir results --n_hop 1 --min_dec_steps 130python test_CGSum.py --visible_gpu 0 --model_dir save_models/CGSum_1hop --model_name CGSum_transductive_1hopNbrs.pt --setting transductive --decode_dir decode_path --result_dir results --n_hop 1 --min_dec_steps 140To test our model , remember to replace the pyrouge root set in data_util/utils.py to your own path.
you can also download our trained model to reproduce our results: inductive 1hop, inductive 2hop, transductive 1hop, transductive 2hop
our dataset is retrieved from S2ORC, the implementation of BertSum can refer to PreSumm, thanks for their works.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent