tdwg/dwc-qa

Use of identificationReferences and identificationVerificationStatus for plankton imaging data

albenson-usgs opened this issue · 3 comments

The plankton imaging community in Europe is evaluating the terms identificationReferences and identificationVerificationStatus for use in identifying the software, version of the software, and machine learning algorithm and if the identification has been validated by human, dubious according to human, or predicted by machine, respectively. This is important information for sharing plankton imaging data as downstream users will want to select subsets of data that have been verified by a human separate from those that are only machine identified. Is it valid to use these terms for this purpose?
@PatriciaCabrera

Thank you @albenson-usgs for initiating this.

After discussion with the community, in the best practices for imaging data management we are publishing next month in Ocean Best Practices, we would like to recommend for identificationVerificationStatus:

  1. PredictedByMachine: for identifications generated by an algorithm and not validated by human.
  2. ValidatedByHuman: for identifications generated by an algorithm and verified to be correct by a human

What would be the process to have this revised by TDWG?

Thanks!

@PatriciaCabrera The process for suggesting changes to Darwin Core can be found at https://github.com/tdwg/dwc/blob/master/.github/CONTRIBUTING.md.

@tucotuco Thank you for your answer. I will look into this.