Link Prediction with Mutual Attention for Text-Attributed Networks

Robin Brochier (1, 2) , Adrien Guille (2) , Julien Velcin (2)

1 DSRT - Digital Scientific Research Technology
2 ERIC - Entrepôts, Représentation et Ingénierie des Connaissances

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.

Mots clés: attributed network; natural language processing; representation learning; link prediction

Domaines: Réseaux sociaux et d'information [cs.SI]

Robin Brochier, Adrien Guille, Julien Velcin. Link Prediction with Mutual Attention for Text-Attributed Networks. International Workshop on Deep Learning for Graphs and Structured Data Embedding, May 2019, San Francisco, United States. ⟨10.1145/3308560.3316587⟩. ⟨hal-02057120⟩