Abstract
When answering queries over knowledge graphs, incomplete answers often arise due to the inherent gaps in the underlying data. Addressing this issue, neural methods have emerged as powerful tools to enrich and complete the knowledge graph, transforming fragmented responses into comprehensive answers. But which neural techniques are best suited for this task? How do they integrate with and enhance downstream query processes? This talk dives into the neuro-symbolic query answering paradigm and explores some of its challenges.