Learning German nouns typically requires that you learn the gender along with the noun. There do exist some "rules" but always with the disclaimer that you can't rely on them.
I wanted to test how often you could rely on them. Taking about 90k German nouns and their known genders, I tested them against the ending rules to see which percentage of them adhered to the rule.
Just a fun morning exercise while I learn some German.
To run:
pip3 install -r requirements.txt
python3 adherence.py
Results (as of 02-Jan-25):
Total: 102444, Skipped: 12405, Processed 90039
Feminine endings:
schaft: 98.71% (m: 1.11%, n: 0.18%) - f=534, m=6, n=1, total=541
heit: 98.46% (m: 0.00%, n: 1.54%) - f=510, m=0, n=8, total=518
keit: 100.00% (m: 0.00%, n: 0.00%) - f=999, m=0, n=0, total=999
ung: 99.16% (m: 0.82%, n: 0.02%) - f=5531, m=46, n=1, total=5578
enz: 97.99% (m: 0.80%, n: 1.20%) - f=244, m=2, n=3, total=249
ion: 96.42% (m: 0.94%, n: 2.64%) - f=1645, m=16, n=45, total=1706
tät: 100.00% (m: 0.00%, n: 0.00%) - f=780, m=0, n=0, total=780
ei: 84.30% (m: 7.04%, n: 8.66%) - f=467, m=39, n=48, total=554
ie: 97.20% (m: 1.73%, n: 1.07%) - f=2185, m=39, n=24, total=2248
ik: 95.53% (m: 4.11%, n: 0.36%) - f=790, m=34, n=3, total=827
in: 82.46% (m: 9.52%, n: 8.02%) - f=4771, m=551, n=464, total=5786
Masculine endings:
ismus: 100.00% (f: 0.00%, n: 0.00%) - f=0, m=559, n=0, total=559
ling: 91.81% (f: 1.42%, n: 6.76%) - f=4, m=258, n=19, total=281
and: 57.75% (f: 15.23%, n: 27.02%) - f=115, m=436, n=204, total=755
ant: 84.72% (f: 2.78%, n: 12.50%) - f=6, m=183, n=27, total=216
ent: 47.75% (f: 0.24%, n: 52.01%) - f=1, m=202, n=220, total=423
ist: 94.40% (f: 4.98%, n: 0.62%) - f=24, m=455, n=3, total=482
er: 79.30% (f: 6.47%, n: 14.23%) - f=640, m=7847, n=1408, total=9895
ig: 90.79% (f: 4.61%, n: 4.61%) - f=7, m=138, n=7, total=152
or: 90.13% (f: 0.17%, n: 9.70%) - f=1, m=539, n=58, total=598
Neuter endings:
chen: 88.53% (f: 0.86%, m: 10.61%) - f=8, m=99, n=826, total=933
lein: 98.00% (f: 0.00%, m: 2.00%) - f=0, m=2, n=98, total=100
ment: 95.69% (f: 0.00%, m: 4.31%) - f=0, m=9, n=200, total=209
trum: 100.00% (f: 0.00%, m: 0.00%) - f=0, m=0, n=52, total=52
nis: 77.88% (f: 19.55%, m: 2.56%) - f=61, m=8, n=243, total=312
tum: 91.98% (f: 0.00%, m: 8.02%) - f=0, m=13, n=149, total=162
um: 73.47% (f: 0.29%, m: 26.23%) - f=4, m=356, n=997, total=1357
o: 54.27% (f: 10.58%, m: 35.14%) - f=78, m=259, n=400, total=737
Total adherence:
Feminine: 93.28%
Masculine: 79.46%
Neuter: 76.77%
Total: 86.57%
I submit this work (specifically, my code) into the public domain or CC0, whichever works best for you.
I would advise that it's more appropriate if using the result data to instead adhere to the license of the dataset itself, which is available at gambolputty/german-nouns.