views:

596

answers:

5

A common BibTeX key convention is last name of lead author followed by year (e.g. Knuth1972). Sometimes this choice is problematic:

  • It could lead to key collisions due to authors with same last name or several publications by the same author in one year.
  • It is hard to tell the subject of the publication from its BibTeX key when editing a LaTeX file: how do you know what \cite{Smith1995} refers to?

Are there any better BibTeX key conventions that address the above problems?

+3  A: 

I (and my research group) use three or more characters identifying the authors, then a two digit year, and then a -2 -3 etc if needed. It leads to short keys that are surprisingly easy to recall.

With one author, Smith: Smi:95

Second paper in 1995 by Smith: Smi:95-2

With two authors, Smith and Fish SmF:05

With three authors Smith, Fish, Wigglesworth: SFW:08

And with more authors, one letter per author: ACFB:09.

With papers that have many, many authors: EJetc:03

Peter
What if you have two Smiths? I am afraid I will soon forget which author is Smi:95-2 and which is Smi:95-3...
eleventy
yep, this can happen. however, a typical paper I write has 100-200 references, and it's less of a problem than you'd think... I typically only need the `-2` once or twice, and (as well as I can remember) it's always been for two related papers by the same authors.
Peter
Ah! I have a single database for all references (I use JabRef to keep them categorized). If you have a separate database for each publication I can see how that would be less of a problem.
eleventy
ah yep, makes sense.
Peter
And what about to add same tag, ie. `daQuirm:08-01` to the name of source file? `Relation_between_light_speed_and_thaumic_field;daQuirm:08-01.pdf`
Crowley
+7  A: 

As I've started putting everything into a single master BibTeX database (with BibDesk), I've been having similar problems. My system is more verbose than yours, and a little more ad-hoc. I write something like

<first-author> <year> - <unique>

I don't like using -1, -2 for papers because of the likelihood of very many authors with the same surname creating works in the same year, so I usually use a keyword such as the journal abbreviation for the <unique> part. If the same author publishes to the same place in a single year I try to add the month or some other meaningful identifier, but it sometimes leads to overly long keys such as:

wallace2009-newyorker-dec

I tend to find this more memorable (albeit harder to type) than a more abbreviated key but of course your mileage may vary.

Will Robertson
I like this approach: looking at a citation of a scientific publication, the conference/journal can help narrow down its topic (e.g. a theoretical CS paper is unlikely to appear in a hardware conference).
eleventy
Do you have to lookup the unique field, or do you tend to remember that?
Charles Stewart
Tend to remember, as long as I'm consistent with my journal/source abbreviations :) Obviously I don't have the name and year memorised for every entry in my bib database, however!
Will Robertson
+1  A: 

I use the format, for example:

KnuthDE86textp

for

Knuth, Donald E., 1985. Tex: The Program. ...

where the textp comes from the first three letters of the first capitalised word in the title, and there are up to three more letters coming from the initials of the following capitalised words in the title. Multiple authors: I key according to the first listed author.

This shortening is adapted from something that Paul Taylor designed for the Imerial College Computer Science bibtex repository, which grew quite large. I've encountered a collisions between author names, but never between keys for works.

I use hand-written code that cobbles together a .bib file from manay .bib files, based on the key. It also supports an alternate-key field (may occur several times), which allows bibtex entries to be accessed by other than the key in pole position in the .bib file.

Charles Stewart
+2  A: 

I paste bibtex entries from Google Scholar and keep the key names.

The key convention for Google Scholar is

<last name of first author> <year> <first word of title>

They are long key names (and hence less key collision). Long key names are no problem because I always rely on the citation completion feature of my editor.

RamyenHead
Wow, thanks a lot, I had no idea you could do that in Google Scholar! I wish they didn't keep this option in the obscure preferences page.
eleventy
A: 

I've used a system like RamyenHead's since I was a was graduate student in the 1970s, and it hasn't failed me yet. The key has three parts: author(s), year, and title. In the author part, capitalize the first letters of names. Use the first three letters of the primary author's last name, followed by the first letter of the author's first name. If there are multiple authors, add the first letters of the other authors' last names in the sequence they are listed on the publication. Then add the last two digits of the year of publication followed by the first non-trivial letter in the title, capitalized.

Thus, we'd have KnuD85T for the Knuth book and KerBP76S for

Kernighan, Brian W., and P. J. Plauger. 1976. Software Tools. Addison-Wesley.

This system takes care of almost all multiple publications in a single year, authors with the same last name, etc., it does not depend on the order in which references are added to the repository, and yet the key itself contains a good deal of info about the reference.

Swamp Man