Automated DOI and journal reference updates from publishers

arXiv collaborates with Inspire (formerly SPIRES) to automatically update arXiv metadata with the DOI and journal references of published versions. In 2006 this facility was extended to accept feeds of DOI and journal reference metadata associated with arXiv articles from publishers and other trusted parties. We believe that linkage between arXiv and publisher versions of articles is of benefit to both arXiv users and publishers. Prospective users should contact arXiv administrators to discuss possible use of this facility, after making the XML feed available on their site for ingestion.

DOI and journal reference update data format

The update information uses a simple XML format to convey DOI and/or journal reference information for a published version of the arXiv article. The XML should use the following schema:

http://arxiv.org/schemas/doi_feed.xsd

and there are some example documents are at:

The body of the feed comprises <article> elements which must have a preprint_id attribute and either of both of the doi or journal_ref attributes:

preprint_id - The identifier of the arXiv article that the doi and/or journal_ref are associated with (e.g. arXiv:1212.1212 or arXiv:hep-th/9901001).
doi - One or more DOIs associated with the article. Multiple DOIs should be separated with spaces. There is no facility to qualify one as the main article and another as an erratum or such hut it is helpful is the main article DOI is first (e.g. 10.1016/S0550-3213(01)00405-9, see also DOI metadata notes).
journal_ref - One or more full bibliographic references to a published version of the article. Multiple references should be separated with a semi-colon and a space. Addition human-readable qualifiers such as "Erratum" may be used (e.g. Phys.Lett. B541 (2002) 273-280; Erratum-ibid. B562 (2003) 367, see also journal reference metadata notes).

The XML format has a few historical oddities. The identifier attribute on the root element is used to check that we have the feed we expect and so should not change for a given publisher/source. The contents of the <date> element are not used during updates. It is useful for human debugging however, and should be set to the date the feed was generated.

Update protocol

After establishing that an update source has reliable and correctly formatted metadata we usually arrange to poll a particular URL on a weekly or monthly schedule. Updates that appear in more than one polling are ignored. For very small feeds (up to 1000 entries say) it is reasonable to simply append to a growing list of updates. For most feeds only recent updates should be included. We recommend that two polling-periods worth of updates are included so that updates won't be missed if something goes awry with one polling cycle.

Problems, errors...

Updates are automatic so it is vital that feeds contain reliable data. Details of errors or other problem should be communicated to arXiv administrators as soon as possible.