Replace latexcodec with pylatexenc, using braces-all mode

[mbollmann]

Cleaner duplicate of #4279 going directly into master, in order to see if this would address the current issues with the BibTeX encoding, see #4280.

This PR tests switching to pylatexenc for LaTeX-encoding strings, which is recommended by the latexcodec documentation and is also faster in my testing. I used the "braces-all" mode of pylatexenc, which should hopefully address #4280.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 92.55%. Comparing base (d4b57ed) to head (716f75b).
Report is 3 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #4284      +/-   ##
==========================================
+ Coverage   92.54%   92.55%   +0.01%     
==========================================
  Files          32       32              
  Lines        2294     2298       +4     
==========================================
+ Hits         2123     2127       +4     
  Misses        171      171              

Files with missing lines	Coverage Δ
python/acl_anthology/utils/latex.py	100.00% <100.00%> (ø)

[github-actions]

Build successful. Some useful links:

• Complete site preview: https://preview.aclanthology.org/fix-bibtex-encoding
• Potential volumes of interest: 2023.finnlp-1, 2023.finnlp-2, 2023.mtsummit-research, 2023.mtsummit-users, 2023.mtsummit-coco4mt, 2024.acl-long, 2024.acl-short, 2024.acl-demos, 2024.acl-srw, 2024.acl-tutorials, 2024.cl4health-1, 2024.dravidianlangtech-1, 2024.findings-eacl, 2024.findings-naacl, 2024.findings-acl, 2024.findings-emnlp, 2024.iwclul-1, 2024.sighan-1, 2024.teachingnlp-1, C82-1, C82-2, W18-1, W18-2, W18-3, W18-5, W18-6, W18-7, W18-8, W18-9, W18-10, W18-11, W18-12, W18-13, W18-14, W18-15, W18-16, W18-17, W18-18, W18-19, W18-20, W18-21, W18-22, W18-23, W18-24, W18-25, W18-26, W18-27, W18-28, W18-29, W18-30, (plus 40 more...)

This preview will be removed when the branch is merged.

[mbollmann]

@danielgildea @nschneid I believe this fixes the issues with the BibTeX-encoding. Could you maybe have a look at the preview branch to spot-check if you see any issues with the way it encodes accents in BibTeX?

[nschneid]

How to check the preview? BibTeX isn't generated for previews right?

[mbollmann]

It is generated for the first three papers of each volume. Those are also compiled in the anthology.bib.gz.

[danielgildea]

Looks great, thank you!

[nschneid]

https://preview.aclanthology.org/fix-bibtex-encoding/2023.cl-2.3/ Lu{\'\i}sa - extra backslash?

https://preview.aclanthology.org/fix-bibtex-encoding/2020.cl-3.3/ is a good test case. Also has {\'\i}.

[mbollmann]

preview.aclanthology.org/fix-bibtex-encoding/2023.cl-2.3 Lu{\'\i}sa - extra backslash?

preview.aclanthology.org/fix-bibtex-encoding/2020.cl-3.3 is a good test case. Also has {\'\i}.

\i is a dotless i, but maybe we shouldn’t have this:

Older versions of LaTeX would not remove the dot on top of the i and j letters when adding a diacritic. To correct this, one had to use the dotless version of these letters, by typing \i and \j. For example:

\^{\i} should be used for i-circumflex î;
\"{\i} should be used for i-umlaut ï.

However, current versions of LaTeX do not need this anymore (and may, in fact, crash with an error).

—https://en.wikibooks.org/wiki/LaTeX/Special_Characters#Escaped_codes

Interesting that pylatexenc produces this by default...

[nschneid]

Tried compiling a bibliography with {\'\i}, and it gave an error, whereas {\'i} worked.

[mbollmann]

I added conversion rules plus tests for í ì î ï to use the regular "i" instead of \i.

[nschneid]

Great—and what about capitalized equivalents?

[mbollmann]

Great—and what about capitalized equivalents?

Added to a test case now, which already passes.