Identity Links

The Problem With owl:sameAs

The sameAs predicate is intended to connect two identifiers that refer to the same thing. For example, both DBpedia and YAGO have entries for Albert Einstein, and it is very useful to be able to connect these. The OWL standard defines sameAs as holding between two identifiers (URIs) whose referents are the same in all possible respects. However, many sameAs links on the Web were generated using automatic methods that merely guess identity. Additionally, even many human-authored links are not strict identity links because some people believe that some sort of near-identity or strong similarity is enough. For example, people might create sameAs links between a beer brewery and the class of all beer bottles produced by that brewery.

For many applications, it is vital to have accurate identity links . In fact, the very idea of Linked Data, as the name suggests, comes from the fact that information from different sources is linked together and thus can flexibly be combined.

The Solution: Simple Alternatives to owl:sameAs

There is no hope in trying to convince the world to stop publishing inaccurate sameAs links. Instead, has defined a separate predicate for genuine identity (lvont:strictlySameAs). This predicate is formally declared equivalent to owl:sameAs, so applications can still interpret these links as regular sameLinks. However, whenever they see that lvont:strictlySameAs was used, they an know that the link is intended in the strict sense.

Thus, the solution is simple. Whenever a human intends to publish a genuine accurate sameAs link, they should use (abbreviated as lvont:strictlySameAs). For better interoperability, a regular owl:sameAs triple can also be published alongside the lvont:strictlySameAs one.

Additionally, the Ontology, also provides separate predicates for whenever one wants to explicitly represent near-identity. In particular, these are (abbreviated as lvont:nearlySameAs) and (abbreviated as lvont:somewhatSameAs). These two predicates are explicitly left somewhat vague, simply because similarity is a very vague notion.

Frequently Asked Questions

What about skos:exactMatch? A skos:exactMatch link only indicates a "high degree of confidence that two concepts can be used interchangeably across a wide range of information retrieval applications".


More Information

For more detailed information about this issue, please refer to the following technical paper: Gerard de Melo (2013). Not Quite the Same: Identity Constraints for the Web of Linked Data  PDF    BibTeX
In: Proc. AAAI 2013.

Return to main page 2008-2016 Gerard de Melo.   Contact   Data Sources   Legal Information / Imprint