A full paper on Generating descriptions of entity relationships by Nikos Voskarides, Edgar Meij and Maarten de Rijke was accepted for publication at the 39th European Conference on Information Retrieval. The paper will be presented in April at Aberdeen, Scotland.
Abstract: Large-scale knowledge graphs (KGs) store relationships between entities that are increasingly being used to improve the user experience in search applications. The structured nature of the data in KGs is typically not suitable to show to an end user and applications that utilize KGs therefore benefit from human-readable textual descriptions of KG relationships. We present a method that automatically generates textual descriptions of entity relationships by combining textual and KG information. Our method creates sentence templates for a particular relationship and then generates a textual description of a relationship instance by selecting the best template and filling it with appropriate entities. Experimental results show that a supervised variation of our method outperforms other variations as it best captures the semantic similarity between a relationship instance and a template, whilst providing more contextual information.