◆ Semantic Web Catalogue

Problem thread

Same thing, many names

When two datasets name the same thing, how do we say so without claiming more than we mean?

issue
identification
also reasoning
status
open
steps
6, 2004–2026
threads
all problem threads

Linked Data lets anyone mint names, so the same person or place ends up with many URIs. OWL offered owl:sameAs to say two names refer to one thing, and Linked Data publishers used it to connect their datasets. Studies then showed that many of these links meant only 'similar' or 'related', and that following them merges things that are not the same. Wikidata took another route, recording other databases' identifiers as properties rather than declaring identity. In practice Wikidata became the hub: by 2026 it links its items to more than ten thousand other identifier schemes.

Lineage

  1. 10 Feb 2004
    proposes standard ★ seminal
    OWL

    owl:sameAs links an individual to an individual: it states that two URI references refer to the same thing, so everything said about one holds for the other. [src]

  2. 27 Jul 2006
    precedent writing ★ seminal
    Linked Data (Design Issues)

    Four rules for publishing data on the Web. The fourth is to include links to other URIs, so that people can discover more things. [src]

  3. Nov 2010
    critiques writing ★ seminal
    When owl:sameAs Isn't the Same: An Analysis of Identity in Linked Data
    ↳ critiques OWL

    Shows that owl:sameAs links between Linked Data sets are often used to mean 'similar' or 'related' rather than strict identity, and proposes finer-grained alternatives. [src]

  4. 29 Oct 2012
    workaround project ★ seminal
    Wikidata

    Records links to other databases as external-identifier properties, one per authority or database, rather than as identity statements between URIs. By 2024 there were nearly 9,000 such properties. [src]

  5. 2018
    measures writing
    sameAs.cc: The Closure of 500M owl:sameAs Statements

    Collects 558.9 million owl:sameAs statements from a 2015 crawl of the Web of Data and computes their closure: 48.9 million sets of names said to be equivalent. Each link gets an error score, because many join things that are not the same. [src1][src2]

  6. 23 Oct 2026
    measures project
    Wikidata Query Service

    Wikidata has become the hub that links other datasets' names for the same thing: a query on 23 October 2026 counts 10,665 external-identifier properties, each linking Wikidata's items to another registry or database. [src]

Open now

Under each question, the answers people have put forward or built. The catalogue gives no odds on them.

  • How can a link say that two things are the same for some purposes but not for others? [src1][src2]
    owl:sameAs allows no degrees; the 2010 analysis proposed finer-grained alternatives, and a 2019 survey reviews the approaches since.
    • Use weaker links than owl:sameAs. SKOS
      SKOS offers exactMatch and closeMatch, which say two concepts match without claiming that everything true of one is true of the other.[src]
  • How do you find the wrong identity links in data you do not control? [src1][src2]
    sameAs.cc scores each link from the structure of the whole network of identity links, since a single bad link can merge large groups of names.
    • Find suspect links from the shape of the identity network. sameAs.cc: The Closure of 500M owl:sameAs Statements
      The sameAs.cc dataset gives each of about 559 million owl:sameAs links an error score computed from the structure of the network, so suspect links can be reviewed first.[src]

How to contribute