Directed Criteria Citation Recommendation and Ranking Through Link Prediction

TL;DR

This model uses transformer-based graph embeddings to encode the meaning of each document, presented as a node within a citation network, and shows that the semantic representations that the model generates can outperform other content-based methods in recommendation and ranking tasks.

Abstract

We explore link prediction as a proxy for automatically surfacing documents from existing literature that might be topically or contextually relevant to a new document. Our model uses transformer-based graph embeddings to encode the meaning of each document, presented as a node within a citation network. We show that the semantic representations that our model generates can outperform other content-based methods in recommendation and ranking tasks. This provides a holistic approach to exploring citation graphs in domains where it is critical that these documents properly cite each other, so as to minimize the possibility of any inconsistencies

Figures (4)

• Figure 1: Augmented Transformer with Learned Residual.
• Figure 2: Embeddings, Colored by Subject Domain
• Figure 3: True Cross-Reference Matrix Organized by Subject Domain
• Figure 4: Predicted Cross-Reference Matrix Organized by Subject Domain, Threshold set at 50%