Publishing research data

Research data publication allows your data to be reused by other researchers e.g. to validate your research or to carry out follow-on research. To that end, a suitable data publication host will allow your data to be discovered (e.g. by publishing metadata) and will be publicly accessible (i.e. on the internet).

Research data can be published on the internet through:
  • project web sites
  • research group web-sites
  • generic web archives (e.g. archive.org)
  • research data sites (e.g. figshare)
  • more general open access research hosts (e.g. f1000 Research)
  • thematic repositories dedicated to a specific discipline / subject area - sadly there is no sign of an appropriate repository for digital music and audio research
  • institutional repositories dedicated to research from a specific organisation (e.g. QMUL have a repository through which Green open access copies of papers by QM research staff can be published).
  • supplementary materials attached to journal articles

An appropriate license should be granted to allow other researchers to use your research data.

Within the Centre for Digital Music, we now have a research data repository for publishing research data outputs from the group. Publishing data though the C4DM repository gives a single point for publishing C4DM data on the internet without relying on (possibly ephemeral) project-specific web-sites. Other repositories that may be of interest to researchers are listed here.

If the web-site through which the data is published is also to be the long-term archive for you data, then you should check that the meets the criteria for an archival storage system. Note that although data will be written to the host irregularly, it is expected that published data will be accessed more frequently than archived data making offline storage unsuitable.

If an external publisher is used for your research data, you should check the Terms and Conditions e.g. to see whether copyright on the data is transferred to the publisher and to check for how long they will publish your data.

If data is published through a publisher or repository, then it may also be held on institutional storage as long as the publisher's license is followed, which might e.g. require that there is a link back to the publisher from the institutional repository. Publishing under a Creative Commons license makes this easy.

If data is available in multiple places, different versions of the data might arise (e.g. changes between dates uploaded, data corruption). You should therefore make it easy to identify which specific version of the data is correct by publishing a digital fingerprint (e.g. a MD5 hash). MD5 fingerprints can be generated in Windows using MD5summer, in Linux with the Gnu md5sum utility and on Max OS X using md5 or openssl

Persistent IDs for data

In order to ensure ongoing access to your data, should look to acquire a persistent ID for your dataset. However, persistence is a continuum with some IDs more persistent than others. DOIs and handles are designed to be persistent in the long term, allowing a unique identifier to be redirected to the current location of your dataset - if the dataset moves, the DOI/handle can be pointed at the new location. Repositories and research data sites may provide DOIs for data submitted to them. Institutional URLs may be persistent if the institution makes a policy decision to make them so. Other URLs may change when web-sites are revamped making the published URL for your data return a "404 Not Found" message.

Persistent IDs are useful for referencing datasets, and are particularly handy if they are short. Long or ugly DOIs can be shortened using the ShortDOI service.

And more repositories