Bo Xiong wrote this awarded paper, Faithful Embeddings for EL++ Knowledge Bases, under the supervision of Profesor Steffen Staab, and with the collaboration of Motjaba Nayyeri, also from the Analytic Computing department, Trung-Kien Tran, from the Bosch Center for Artificial Intelligence, and Nico Potyka, from the London Imperial College. This paper is also part of the KnowGraphs project, whose main goal is to scale knowledge graphs to be accessible to a wide audience of users across multiple domains, including companies of all sizes and even end users.
So far, the learning representation of symbolic knowledge bases have been limited to the data-level, or have suffered from inherent limitations when dealing with concept-level knowledge. Existing learning representations could not faithfully model the logical structure of a knowledge base. In their paper, Xiong et al. presented BoxEl, a geometric KB embedding approach that allows for better capturing the logical structure of EL++ knowledge bases. Experimental results on plausible subsumption reasoning and a real-world application dealing with protein-protein interactions show that BoxEL outperforms traditional knowledge graph embedding methods.
The International Semantic Web Conference is the most relevant conference on the topic of semantic technologies and knowledge graphs and is ranked as an A conference by CORE. The Best Paper Award and the Best Student Research Paper Award were selected among three nominations out of a pool of 153 submissions to the research track of the conference.
Find the paper at https://doi.org/10.48550/arXiv.2201.09919.
