6. Dealing with duplicationΒΆ

Duplication between identifier records is not a new problem and is common to many applications (e.g. bibliographic, medical records). While PIDINST identifiers are considered globally persistent it is accepted that duplication may occur particularly where institutions loan instruments to other organisations or provide access to specialised facilities (e.g. large scale synchrotrons, medical laboratories, computational facilities). Such duplication may lead to inaccurate statistics or reporting about instrument assets.

It is recommended that owners of instruments try to employ workflows and procedures that avoid duplication in the first instance. Where this has not been possible, it is recommended that instrument owners employ deduplication, effectively merging duplicate records into one representative record by ensuring links between them. This can be achieved using the PIDINST metadata schema relatedIdentifier property with a relationType attribute IsIdenticalTo as shown in Snippet 6.1.

<relatedIdentifiers>
   <relatedIdentifier relatedIdentifierType="DOI" relationType="IsIdenticalTo">10.4232/10.CPoS-2013-02en</relatedIdentifier>
</relatedIdentifiers>
[{
   "RelatedIdentifier":{
      "RelatedIdentifierValue":"10.4232/10.CPoS-2013-02en",
      "RelatedIdentifierType": "DOI",
      "relationType":"IsIdenticalTo"
      }
  }]

Snippet 6.1: Merging duplicate instrument PID records using (1) XML and (2) JSON

Recent advances in technologies are expanding to algorithms that automatically detect and resolve deduplication. While such methodologies have been known to improve the efficiency of detection in large collections such as Google Scholar or OpenAire Research Graph, algorithms may be limited by heterogeneous representations for example, by the use of differing semantics. While automatic detection is encouraged, the PIDINST schema is designed to complement multidisciplinary best practices for property values and many properties allow for soft-typing, giving users the ability to use values of their choice, such as free text or domain-specific standards.