Publishing DOIs instead of traditional references?

Question

Today's world is electronic, there is vision of semantic web and yet we still submit references in numerous versions of various journals. In medicine, many journals have special rules about after how many authors you can put et. al. etc etc.

It is tedious to reformat references (even with EndNote or other reference managers) to suit a particular journal.

Are there any pioneer journals or initiatives to simply stop submitting references and just list a DOI instead (if one exist), and to "traditional refences only if DOI does not exist.)

LATER EDIT: In fact, people already try to do the opposite - convert PDF into list of DOIs. http://labs.crossref.org/pdfextract/

LATER EDIT: The receiving journal, after getting the article, would expand the DOIs into traditional references in the exact format the journals wants the references in the print version. Different journals would expand the DOI as preferred. A service at http://www.crossref.org/guestquery can expand the DOI (e.g., 1136/bmj.d7373 into XML with info such as:

<?xml version="1.0" encoding="UTF-8"?>
<doi_records>
  <doi_record key="key" owner="10.1136" timestamp="2012-01-03 23:32:36">
    <crossref>
      <journal>
        <journal_metadata language="en">
          <full_title>BMJ</full_title>
          <abbrev_title>BMJ</abbrev_title>
          <issn media_type="print">0959-8138</issn>
          <issn media_type="electronic">1468-5833</issn>
        </journal_metadata>
        <journal_issue>
          <publication_date media_type="online">
            <month>01</month>
            <day>03</day>
            <year>2012</year>
          </publication_date>
          <publication_date media_type="print">
            <month>01</month>
            <day>03</day>
            <year>2011</year>
          </publication_date>
          <journal_volume>
            <volume>344</volume>
          </journal_volume>
          <issue>jan03 1</issue>
        </journal_issue>
        <journal_article publication_type="full_text">
          <titles>
            <title>Compliance with mandatory reporting of clinical trial results on ClinicalTrials.gov: cross sectional study</title>
          </titles>
          <contributors>
            <person_name sequence="first" contributor_role="author">
              <given_name>A. P.</given_name>
              <surname>Prayle</surname>
            </person_name>
            <person_name sequence="additional" contributor_role="author">
              <given_name>M. N.</given_name>
              <surname>Hurley</surname>
            </person_name>
            <person_name sequence="additional" contributor_role="author">
              <given_name>A. R.</given_name>
              <surname>Smyth</surname>
            </person_name>
          </contributors>
          <publication_date media_type="online">
            <month>01</month>
            <day>03</day>
            <year>2012</year>
          </publication_date>
          <publication_date media_type="print">
            <month>01</month>
            <day>03</day>
            <year>2012</year>
          </publication_date>
          <pages>
            <first_page>d7373</first_page>
            <last_page>d7373</last_page>
          </pages>
          <publisher_item>
            <item_number item_number_type="sequence-number">bmj.d7373</item_number>
          </publisher_item>
          <doi_data>
            <doi>10.1136/bmj.d7373</doi>
            <timestamp>20120103162404</timestamp>
            <resource>http://www.bmj.com/cgi/doi/10.1136/bmj.d7373</resource>
          </doi_data>
        </journal_article>
      </journal>
    </crossref>
  </doi_record>
</doi_records>

Answer 1

It's not the horrible suggestion that many people seem to think it is -- there have been some complaints about miscounts in citation counts because of improper parsing of references; mostly because of some journals allowing non-standard abbreviations. (does A&A mean Astronomy and Astrophysics or Arts & Architecture? Many citation parsers don't use the context and have alternate rules for what journal the citation came from.

This could also bring up some other problems -- if someone's published a pre-print in arXiv ... but the final publication which may be substantially different. Very few people cite the arXiv paper directly, but it might actually be more appropriate than citing the final published version in some circumstances. (although, it would also show that perhaps those portions that you're citing didn't stand up under peer-review); you couldn't cite the arXiv via DOI, only via bibcode.

...

Now, in most situations, the reference list for some journals is handled differently from the rest of the paper. In talking with the editor for Solar Physics, when I talked to him regarding a proposal for data citation, he mentioned that a problem was that giving a second URL to provide specifics would be an issue, and I believe he hinted that it was the publisher and not something locally developed.

Ideally, this whole thing would be integrated into the submission process, and then the reviewers would have the list expanded so that they can do their work. (and you'd probably want the submitter to verify the expansions) ... but this of course will take time from the publishers, who aren't exactly known to be nimble for the most part.

...

What I'd suggest, so that you can get the benefits that you want, without waiting for the journals to make any changes, is to look into reference manager software, which let you collect your references in a database, with notes, and can then generate a reference list using the appropriate format. I've heard good things about a number of them (but never had a need for this purpose (the ones I've used were for shared project bibliographies). I'd look to see which ones can generate whatever citation style is appropriate for the journal(s) you submit to, or ask around to see if any of the people you collaborate might use one (as using the same one could be of benefit when collaborating).

Answer 2

This is a horrible idea. DOIs are not meant for human consumption; bibliographies are.

You cannot assume that your readers are always connected to the internet while reading your paper, even if that paper is published only electronically. They might be reading a paper printout. They might be reading the paper on an airplane. They might be describing the results to students in a seminar, or to a colleague over the phone. The campus network might be down, or overloaded by students google-studying for finals. Their cellular data carrier might be unreliable.

Even so, the reader might have access to your cited papers by other means, like (gasp) the actual library. Or they might be interested in when your references were published, or in which venues, or by whom. Or they might want to track citations to other papers which are not identified by DOIs elsewhere. Or they might want the raw bibliographic information for some other reason which you and I can't think of, but which they've learned to expect because every other publication includes it.

tl;dr: Stop whining and just type the stupid page numbers already.

Answer 3

I agree with JeffE and Users 17670, but there's another fundamental reason why DOIs alone are problematic: they don't have enough redundancy to correct errors. If you have a typo in your DOI, then it will be very difficult to determine what you meant to refer to. Of course, most DOIs are copied and pasted, and it's not hard to check that they work by pasting them into http://dx.doi.org, so I'm sure the error rate would be very low. However, it would not be zero. (I once found a typo in the DOI listed on a publisher's web page. I have no idea how it got there, but it did.)

Answer 4

Personally, I try to "sneak in" DOIs into reference lists even if the journal's instructions are for short citation formats. Just as I try to sneak in the title of the paper.

• As reader and reviewer, I find it horrible to have no clue at all what the paper by J. Doe et al., Some Jounal, 42, 5678-91 is about. Particularly, when seeking a particular kind of information. That is for the title.

• As reader and reviewer, I find it very convenient to just click on the link to get to the paper. So please, do not only spell out the DOI, but use its inherent hyperlinking capability.

• My experience is that broken citations are fairly common already with the traditional "Author, journal, volume, pages" format. Usually it is possible to arrive at the paper with some effort. Having the title is then the most useful information to get to the paper. Having only a DOI with typo pretty much prevents any possible correction.

• Conclusion: I think both title and DOI are worth the additional ink.

• Even if the journal insists on their more compressed format, I'd like to have at least the reviewers the benefit of immediately seeing what papers I refer to.

• I've never had a reviewer complaining about this information in the references.

• I'm using mostly biblatex, which makes it easy to automatically have both title and the DOI (already properly linked) output together with the rest of the information. No effort at all.

• I don't care if you prefer to store your literature database as DOI only, and expand the DOI list when writing the paper. However, personally I perfer to have a searchable base, and more meaningful citation keys (I use mostly AuthorYEAR Bibtex keys, for books the title), so I can remember them more easily when writing.

Answer 5

People already see this as a problem and try to go from PDF to a list of DOIs. See the project here: http://labs.crossref.org/pdfextract