Link Prediction with Mutual Attention for Text-Attributed Networks
This work addresses link prediction for researchers using citation networks, but it is incremental as it adapts existing attention mechanisms to a specific domain.
The paper tackles link prediction in text-attributed networks by learning document similarity from network topology using a mutual attention mechanism, showing preliminary results on citation datasets with improved performance.
In this extended abstract, we present an algorithm that learns a similarity measure between documents from the network topology of a structured corpus. We leverage the Scaled Dot-Product Attention, a recently proposed attention mechanism, to design a mutual attention mechanism between pairs of documents. To train its parameters, we use the network links as supervision. We provide preliminary experiment results with a citation dataset on two prediction tasks, demonstrating the capacity of our model to learn a meaningful textual similarity.