Stats on new authors

[nschneid]

It occurred to me to check how many of the authors in our large conferences are new to the community, as measured by not having any Anthology publications in a prior year. Of course author pages are imperfect so these numbers are ballpark figures, but they're pretty striking. Counting only main conference long/short papers:

• ACL 2024: 4209 unique authors, 1318 first-time (31%), 2891 repeat from a previous year
• EMNLP 2024: 5460 unique authors, 2118 first-time (39%), 3342 repeat from a previous year
• ACL 2025: 8072 unique authors, 3037 first-time (38%), 5035 repeat from a previous year
• EMNLP 2025: 8255 unique authors, 3424 first-time (41%), 4831 repeat from a previous year
• EACL 2026: 2074 unique authors, 579 first-time (28%), 1495 repeat from a previous year

EACL is a smaller conference, of course; I included it because we have an instance from this year. Comparing with some older conferences:

• ACL 2020: 2509 unique authors, 724 first-time (29%), 1785 repeat from a previous year
• ACL 2015: 881 unique authors, 235 first-time (27%), 646 repeat from a previous year

I guess the 40%ish rate of newcomers we are now seeing at ACL and EMNLP, as opposed to 28%ish percent earlier, indicates the growth of the field continues to accelerate.

Code:

from itertools import chain
allpapers = chain(anthology.get('2026.eacl-long').papers(), anthology.get('2026.eacl-short').papers())
allauthors = []
repeat = 0

for paper in allpapers:
        for ns in paper.authors:
            author = ns.resolve()
            if author not in allauthors:
                allauthors.append(author)

for author in allauthors:
        if any(int(paper2.year)<2026 for paper2 in author.papers()):
            repeat +=1

print(f'EACL 2026: {len(allauthors)} unique authors, {len(allauthors)-repeat} first-time, {repeat} repeat from a previous year')