The code and data for citation intent prediction used in the paper:
GraphCite: Citation Intent Classification in Scientific Publications via Graph Embeddings,
accepted at the workshop Sci-k 2022, in conjunction with TheWebConf 2022:
https://sci-k.github.io/2022/
We provide models for doing citation intent prediction on ACL-ARC and SciCite datasets. To run the code please install the libraries specified in requirements.txt
The code can be run as follows:
usage: graphcite.py [-h] [-b BATCH_SIZE] [-e EPOCHS] [-s SEED] [-lr LEARNING_RATE] [-a AUTHORS] [-v VENUES] dataset model
The following arguments are required: dataset, model
The default values are:
- batch size 16
- epochs 10
- seed 0
- learning rate 2r-5
- adding authors and venues to citation graphs is by default set to false
The parameter dataset can take two values: acl-arc or scicite
The parameter model can take the following values:
CitationBERT - fine tuning BERT on citation prediction
CitationGAT - training GAT on citation prediction
CitationBERTGAT - our GraphCite model
CitationMLP - a simple MLP layer on top of the titles embeddings
To cite this work:
@inproceedings{10.1145/3487553.3524657, author = {Berrebbi, Dan and Huynh, Nicolas and Balalau, Oana}, title = {GraphCite: Citation Intent Classification in Scientific Publications via Graph Embeddings}, year = {2022}, isbn = {9781450391306}, publisher = {Association for Computing Machinery}, address = {New York, NY, USA}, url = {https://doi.org/10.1145/3487553.3524657}, doi = {10.1145/3487553.3524657}, abstract = {Citations are crucial in scientific works as they help position a new publication. Each citation carries a particular intent, for example, to highlight the importance of a problem or to compare against results provided by another method. The authors’ intent when making a new citation has been studied to understand the evolution of a field over time or to make recommendations for further citations. In this work, we address the task of citation intent prediction from a new perspective. In addition to textual clues present in the citation phrase, we also consider the citation graph, leveraging high-level information of citation patterns. In this novel setting, we perform a thorough experimental evaluation of graph-based models for intent prediction. We show that our model, GraphCite, improves significantly upon models that take into consideration only the citation phrase. Our code is available online1.}, booktitle = {Companion Proceedings of the Web Conference 2022}, pages = {779–783}, numpages = {5}, keywords = {citation intent classification, graph neural network}, location = {Virtual Event, Lyon, France}, series = {WWW '22} }