Empirical Studies of Context at Scale: The Case of Equality Reasoning or: How Leibniz Got it Wrong

摘要

The rise of very large linked open datasets has allowed us over the past few years to study the structure of knowledge graphs not only in theory, but also empirically at very large scale. I will report on a number of studies that all have empirically analysed the role of context in equality reasoning in linked open data, encoded in the owl: sameAs predicate. All of these studies show that the standard formal semantics of equality does not suffice in practical settings, and is simply ignored and violated at a large scale. At the same time, we can show that different notions of context are very useful in making sense of what users choose to do in practice, and that within local contexts, a sensible semantics for equality reasoning does emerge.