ArXiv: Academic Network
The ArXiv dataset is a heterogeneous academic network constructed from arXiv CS and the Microsoft Academic Graph [1]. It contains three types of entities—papers (81,634 nodes), authors (127,590 nodes), and fields of study (FoS, 21,887 nodes)—as well as three types of directed relations connecting three types of entities—a paper “cites" another paper, an author "writes" a paper, and a paper “has a topic of" a field of study. Papers are associated with their titles and abstracts, and all the other types of entities are textless. We concatenate the paper title and abstract and pass it to a MiniLM sentence encoder, generating a 384-dimensional feature vector for each paper node. Papers are categorized into 40 subject areas of arXiv CS papers, such as cs.LG, cs.DB, and cs.OS. In addition, each paper is associated with the year of publication.
Prediction task. The task is to predict the 40 subject areas of arXiv Computer Science papers, which are manually assigned by the paper's authors and arXiv moderators. Given the rapid growth of scientific publications, doubling every 12 years over the past century \citep{dong2017century}, automating the classification of papers into relevant areas is of significant practical value. Formally, the task is formulated as a 40-class classification problem, with evaluation metrics including the Micro-F1 and Macro-F1 scores.
Dataset splitting. For the dataset split, we use a temporal approach. The model is trained on papers published up until 2017, validated on those from 2018, and tested on papers published from 2019 onward. This temporal split reflects a practical scenario of helping the authors and moderators annotate the subject areas of the newly-published arXiv papers.
Dataset statistics.
| # Nodes | # Edges | # Classes | # Splits |
|---|---|---|---|
| 231,111 | 2,075,692 | 40 | Train: 47,084 |
| Paper: 81,634 | Paper-Paper: 1,019,624 | Valid: 18,170 | |
| Author: 127,590 | Paper-Author: 300,233 | Test: 16,380 | |
| FoS: 21,887 | Paper-FoS: 755,835 |
[1] Wang, K., Shen, Z., Huang, C., Wu, C. H., Dong, Y., & Kanakia, A. (2020). Microsoft academic graph: When experts are not enough. Quantitative Science Studies, 1(1), 396-413.
Dataset Construction
- Install ogb and dgl
- Download source data Arxiv.csv
- Download PAuAf_CS_20190919.tsv and PF_CS_20190919.tsv
- Run
graph_builder.ipynb