How to Capitalize a Title in {BibTeX}

by Jason Eisner (2020-)

You need to proofread the typeset PDF version of your bibliography. Of course a sloppy bibliography doesn't reduce the scientific value of your paper, but it looks unprofessional, just as typos do.

• Check that the substantive information is accurate.
• Check that capitalization is correct, especially in article titles.
• Check that author name formatting is correct and consistent, especially initials and unusual names.
• Include URLs or DOIs wherever possible, especially if your .bst file will hyperlink the title to the URL.
• Check that you were consistent across entries in the level of detail you provided and in your abbreviation conventions.
• Prefer the published version of a paper to an arXiv preprint.

Capitalization is usually the biggest problem. Do you see oddly capitalized titles like "How to capitalize a title in bibtex"? You need to fix the entry to protect "BibTeX" from getting lowercased:

    title = {How to Capitalize a Title in {BibTeX}}

You can't just type the title the way you want it to look, since BibTeX may change it to something else, depending on the bibliography style (e.g., \bibliographystyle{acl_natbib}). You are supposed to write the .bib file entries so that they'll look correct with any bibliography style (.bst file).

When BibTeX generates a LaTeX bibliography from your .bib file, it will use the .bst file to format it. Different publication venues have different capitalization conventions and will therefore use different .bst files. So the title entry above, which is correct, might in theory appear as any of the following:

• How to Capitalize a Title in BibTeX
• How to Capitalize a Title in BibTeX    ← ACL style formats book titles like this
• How to capitalize a title in BibTeX     ← ACL style formats paper titles like this
• HOW TO CAPITALIZE A TITLE IN BibTeX

One of these days, someone (me?) will build a nice .bib cleaner-upper. In the meantime, you can follow the instructions below to get your .bib entries right the first time.

Definitions

• uppercase = ABC-DEF GHI
• lowercase = abc-def ghi
• titlecase = Abc-Def Ghi

Part 1: How should you format the title printed at the top of the PDF paper?

These rules can also be used for titles of figures and tables.

1. First write the title in LaTeX just as if it were a phrase in the body of the paper:

   a named-entity recognizer to fix spelling: the case of Arabic OCR on the iPhone

   You should use uppercase letters just when the word’s dictionary entry does so: Arabic, Bayesian, ChatGPT, CRF, HMMs, iPhone.

   Use TeX formatting for special characters and formulas:

   the $\ell(\varepsilon)$ function of Poincar\'{e}

2. Now change all lowercase words to titlecase, except for "minor" words. The definition of "minor" words varies, but APA suggests

   • the basic determiners: a an the
   • conjunctions of ≤ 3 letters: e.g., and as but for if nor or so yet ...
   • prepositions of ≤ 3 letters: e.g., as at by for in of off on per to up via ...
   and many people add
   • the infinitive marker to

   Now you have:

   a Named-Entity Recognizer to Fix Spelling: the Case of Arabic OCR on the iPhone

3. Change the first and last words to titlecase as well, even if they are "minor" words.

   If the title consists of multiple parts separated by any of the punctuation marks : --- -- ? ! ., then use titlecase for the first and last words of each part.

   Exception: Never change a word such as iPhone that inherently starts with a lowercase letter.

   This gives your final title:

   A Named-Entity Recognizer to Fix Spelling: The Case of Arabic OCR on the iPhone

Part 2: How should you format that title in BibTeX?

These rules should be used for BibTeX fields such as title, booktitle, and journal.

1. Start by copying the title exactly from step 1c above:

   title = {A Named-Entity Recognizer to Fix Spelling: The Case of Arabic OCR on the iPhone}

2. Use curly braces to protect any material whose case must be preserved for meaning. These are the inherently uppercase or lowercase letters. (The inherently uppercase letters were already uppercase at step 1a.)

   title = {A Named-Entity Recognizer to Fix Spelling: The Case of {A}rabic {OCR} on the {iPhone}}

   Also, if you used titlecase following a punctuation mark other than colon, you should protect that initial uppercase letter:

   title = {A Named-Entity Recognizer to Fix Spelling---{T}he Case of {A}rabic {OCR} on the {iPhone}}

   This is because most .bst files know to treat the word following colon like the first word of the title, but unfortunately they do not extend this courtesy to the punctuation marks --- -- ? ! . that more rarely appear in titles.

   Leave other material unprotected, so that BibTeX can change the capitalization to match the bibliographic conventions of the publication venue that supplied the .bst file.

How about other tricky BibTeX Fields?

url = {https://ui.adsabs.harvard.edu/abs/1948PhRv...73..803A/abstract} or doi = {10.1103/PhysRev.73.803} is very helpful. Include a url or doi field whenever you can. A good .bst file such as acl_natbib will use this to turn the paper title into a convenient hyperlink, rather than printing an ugly URL or DOI as part of the bib entry. If necessary, such a .bst file can be produced with the help of urlbst, using the --inlinelinks option.

pages = {123--134} uses an en-dash for a page range.

month = apr gives the month name as a symbolic constant, without quotes. It is the .bst file's job to translate that into something like Apr, April, Avril (in French), or 04.

author = {Alan Turing and I.~J. Good and Hugh Alexander} is a typical example of an author list. This format is also used for editor. Whether to abbreviate a name is the .bst file's decision, not yours, so you should provide the full name as shown on the paper, using initials only if the paper itself uses them. An initial should always be followed by a period and a space. Two initials in a row are preferably separated by the non-breaking space ~. Names are unfortunately complicated; more guidance is here and here.

journal = {J. Mol. Biol.} may be an alternative to journal = {Journal of Molecular Biology}, but use a consistent style of abbreviation throughout the bibliography.

publisher = {Association for Computational Linguistics} can usually be omitted. I tend to omit publishers for conferences and journals that are well-known to the paper's intended audience, but keep them for more unfamiliar venues and for books.

To suppress unwanted fields from a bibliography entry, without deleting the information, you can change the field names. For example, change publisher, address, month, pages to OPTpublisher, OPTaddress, OPTmonth, OPTpages. (Another option is to modify the .bst file so that it doesn't print those fields.)

How do I BibTeX an arXiv paper?

First, check whether the version you want has been published somewhere. In that case, make an entry for the published version instead, to show that the paper appeared officially in a peer-reviewed forum. Often the abstract page on arXiv will have a field that tells you where to look:

Comments: NAACL-HLT 2021

But if you need to cite the arXiv version, I suggest this:

@article{rasooli-tetrault-2015,
  author    = {Mohammad Sadegh Rasooli and Joel R. Tetreault},
  title     = {{Y}ara Parser: A Fast and Accurate Dependency Parser},
  year      = {2015},
  journal   = {Computing Research Repository},
  volume    = {arXiv:1503.06733},
  url       = {http://arxiv.org/abs/1503.06733},
}

  volume    = {arXiv:1503.06733~},
  number    = {version 2},
  url       = {http://arxiv.org/abs/1503.06733v2},

Where can I find existing BibTeX entries online?

Unfortunately, it is often necessary to correct online entries. They typically get capitalization wrong, and they may have a suboptimal URL. Nonetheless, they can be a good starting point.

• A web search for the paper title together with InProceedings will often turn up a BibTeX entry, if it's a conference paper.

• You can grab a huge bibfile of all ACL-related papers from ACL Anthology.

• You can search

  • DBLP (then click the "download" icon")
  • arXiv (then click "Export BibTex Citation")
  • Semantic Scholar (then click "Cite")
  • Google Scholar (then click "Cite")

• AI will generate BibTeX entries for you, especially for well-known papers. Be sure to check these entries by hand! It's good at the formatting, but it tends to hallucinate some of the bibliographic details. This can get your paper desk-rejected.

Can I use AI to correct capitalization and other formatting in an existing BibTeX file?

Yes, I've written a long prompt for this that works moderately well. I then use a diff tool to review the changes. Perhaps I should post my prompt here (feel free to email me).