When owl:sameAs Isn’t the Same: An Analysis of Identity in Linked Data

Abstract

Abstract. In Linked Data, the use of owl:sameAs is ubiquitous in interlinking data-sets. There is however, ongoing discussion about its use, and potential misuse, particularly with regards to interactions with inference. In fact, owl:sameAs can be viewed as encoding only one point on a scale of similarity, one that is often too strong for many of its current uses. We describe how referentially opaque contexts that do not allow inference exist, and then outline some varieties of referentially-opaque alternatives to owl:sameAs. Finally, we report on an empirical experiment over randomly selected owl:sameAs statements from the Web of data. This theoretical apparatus and experiment shed light upon how owl:sameAs is being used (and misused) on the Web of data.

On the Likelihood of an Equivalence

“…owl:sameAs statements do not always honor this rigorous semantics. For example, [3] reports four very different typologies of use of owl:sameAs they actual found in Linked Data (contextualization, referential opacity, similitude, and reference misplacement). While these different uses appear to be acceptable and sometimes even useful for some applications (e.g.…”

“…For example, [5] points out that rdf:seeAlso is much "weaker" than owl:sameAs; and [3] that the SKOS vocabulary has a number of "matching" predicates that are close in meaning to owl:sameAs without however implying full identity (skos:broadMatch, skos:narrowMatch, skos:closeMatch, etc.). However, their use may be subjective and even if one is tempted to engage with some sort of numerically weighted uncertainty measure of identity, the real hard questions of where precisely will these real values come from [3] arises. Our first question to answer.…”

What Should I Link to? Identifying Relevant Sources and Classes for Data Linking

“…The meaning of the equivalence relation here depends on the identity criterion chosen by the data publisher: e.g., owl:sameAs links or direct reuse of URIs assume that URIs must be strictly interchangeable (see [4] for the analysis of different types of identity). The goal is to identify the subset of relevant repositories {D 1 , .…”

Key Ingredients for Your Next Semantics Elevator Talk

“…-The adopter's data will become part of the active research community which is concerned with analyzing, understanding, improving, interlinking, and using Linked Data for various purposes. With those benefits in mind, it is also important to point out what Linked Data does not deliver [11][12][13].…”

“…-The links in Linked Data are often created ad-hoc with a more-is-better mentality instead of strategies to assess quality, or to maintain and curate already established links. Indeed, many of those links are owl:sameAs links which, however, are usually not meant to carry the formal semantics they would inherit from the Web Ontology Language OWL [11,14]. -The paradigm shift to triples as units of meaning and URIs as global identifiers alone is not sufficient to contribute to the Linked Data cloud.…”