Reference validation

[fred-atherden]

References need to be validated against canonical bibliographic sources (e.g. PubMed, Crossref etc) so that the correct details are included.

User Stories

Author

1. As an author, I want to be able to populate a reference by searching a bibliographic database so that I don't have to manually fill out every field.
2. As an author, I want to be able to use a DOI or PMID to locate the details for a reference so that I can complete references using the minimum possible information.
3. As an author, I want to be able to use a reference's author/title/year details to retrieve the corresponding DOI and/or PMID so that I can complete the reference details.

Production staff

But what if . . . ?

Considerations

• An author should be able to enter, say, first author, year, title and journal and search for the rest of the information, or search using DOI via connections with indexing services like PubMed and Crossref.
• When an article uses author-year citations, the validation tool should not override the a,b,c etc additions to the year field used to distinguish between citations with the same name and year UNLESS the validation results in the year changing.
• eLife is very strict about the order of preference for data. For instance, PubMed has better quality and curated metadata but is less comprehensive than Crossref, so current rules dictate a references is validated against PubMed first and if there is no result, the tool then tries to validate it against Crossref.
• eLife has some rules in place to establish known mistakes in the metadata - for instance, Crossref has only included elocation-ID in their schema since 2018, so many journal first pages are actually elocation IDs. The current tool converts this tag back to elocation-ID for journals known to use them (e.g. PLOS, eLife).
• Some journals have changed name over time and for purposes of neatness in the reference list, eLife overwrites the database output with the preferred version (captured in a configuration list). The best example of this is PLOS, which can be presented as plos, PLOS or PLoS. We normalise to 'PLOS'. In addition, there are some titles that are captured rather oddly in PubMed (e.g. Nature Genetics being recorded as 'Nature. Genetics'. These are also over-ridden with a more stylistically suitable version.
• Many book publishers have been assigning DOIs to books and book chapters. eLife's current tool/process is poor at retrieving these and needs to be improved.
• eLife's current tool does not validate URLs in references. It would be great if the tool could ping a webpage and alert us if an error page is reached.
• eLife has a list of journals with dates before which the journal did not assign DOIs to its content, and where it is known that DOIs have not been back-filled in their archives.

XML requirements

Taken from: https://github.com/elifesciences/TextureRequirements/issues/90

[NickDuf]

It is possible to add a reference using DOI in texture. This feature does not meet eLife’s user stories completely