Under review

Publication links not unique

Adam Rego Johnson 1 year ago updated 1 year ago 7

Hi, the links for a particular publication do not appear to always be unique. The formula seems to be something like "bibbase.org/network/publication/author1-author2-sometitle-yyyy". The problem comes when you have two references with the same authors, title, and year. This happens somewhat often in my database, as I have econ papers that have two versions -- a published paper and a working paper version of that paper (which often contains much more information and data then the published version). On my BibBase page itself, these are shown just fine of course, but if I want to link directly to a reference, the generated URL is the same for both references.

An example can be seen on my page here. Under 2021, both version of the Saez & Zucman paper "A Wealth Tax on Corporations' Stock" both have the same "link" (accessible here) and the link only displays the working paper version.

We use the direct BibBase links to references fairly often in one part of our project in order to link to the original sources we are citing -- one section in particular may be vulnerable to this, as we are analyzing the methodologies of various economic papers. Sometimes we analyze the published version, sometimes the working paper versions given the increased information, and sometimes those two papers would end up having the same link as they are in the same year and with the exact same title and authors.

Is there any way to ensure all links would be unique and thus more stable?

Thank you!


Under review

That's not currently possible, no, unless you are willing to change the title of course. In computer science, e.g., it is not uncommon to have such papers with more information, but they usually indicate this in their title, e.g., by adding "Extended Version" or similar to avoid confusion.

I think we can add support for this, though, by using the bibtex key for disambiguation.

I see. Yes, most often the working papers don't have any indicators in their titles so no dice there (Economics has some unique quirks, as I have discovered).

How could the bibtex key be used here? I can add that if it pulls from the .bib rendering the publication list, I'd imagine it would work fine in my case -- my library's bibtex keys are of course unique, which is why I have actually ended up having to use them as the file names in order to simplify things and avoid problems with PDF file names in relation to this same problem. But given that the publication links are not unique to particular .bib files / publication lists as far as I can tell, I don't see how you could use bibtex keys which may be different and/or overlapping in various libraries. But I am open to any solution of course, thank you.

Our plan is to use the bibtex key in addition to the bibbaseid, just for disambiguation when necessary. I'll keep you posted.

Ah, okay. Thank you!

Hi, do you know when you might have updates to share with me regarding this?

Thank you!

Hi Adam,

I've looked into this and unfortunately the solution we had in mind is a lot more involved then we thought. The issue is that the current bibbaseid (authors-title-year), is used for indexing in the database and is assumed to be unique. This doesn't mean that this solution won't work, but it will require a lot more time to implement and test than anticipated. As a result we don't currently know when we'll have the time to address this, sorry.

For now, I think the only option is to add something to the title like "(working paper)".

Ok, thanks for looking into this, oh well! I'll figure out the best way to implement something like this in the titles in our library. Thank you.