University of Twente Student Theses


Augmenting and Graph Scaling for Context-Aware Citation Recommendations

He, Wenruo (2022) Augmenting and Graph Scaling for Context-Aware Citation Recommendations.

This is the latest version of this item.

[img] PDF
Abstract:Citing sources could be a tedious task when writing scientific papers, and a local citation recommendation system is designed to solve this problem by providing recommendation for each citation context. Inspired by Jeong, Jang & Park who purposed the BERT-GCN model and Bhowmick, Singhal & Wang who improves the BERT-GCN model by adding co-authorship graph, we explored other possible extra features for further improvement. Focusing on author, venue information and centralities, we formed a feature matrix and replaced on the featureless matrix used in GCN. Due to the lack of benchmark datasets, we constructed a local citation recommendation targeted dataset based on Cord-19 dataset, and together with the public available FullTextPeerRead dataset, we evaluated our models. To tackle the limitation of GCN on large graphs, we replace GCN with SIGN on the large CORD19 dataset. Finally, we evaluated our models on MAP, MRR and Recall@k and compared them to the baseline models. The results showed that the uncleaned and possibly bad choice of extra features may be the fatal flaw. However, SIGN embeddings still shows competitive performances, which indicates their ability of processing large graphs.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science
Programme:Computer Science MSc (60300)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page