French Noun Gender Dataset

A single CSV file listing 141,872 French nouns and the grammatical gender of each one. It is the same data that powers the lookup on this site, merged from two open dictionaries and free to reuse.

Download the CSV Read the notice

What is in it

A plain UTF-8 CSV with two columns:

Column	Contents
`word`	The noun as it is written, accents included.
`gender`	`masculine`, `feminine`, or `epicene` for a word that takes either gender or where the two sources disagreed.

Rows are sorted in French dictionary order. The file is about 4 MB.

Licence

The dataset contains data from DELA, so the file as a whole is offered under the Lesser General Public License For Linguistic Resources (LGPL-LR), the same licence DELA uses. In plain terms:

You may use it for anything, including commercial work.
You may redistribute it, including a changed or filtered version, as long as the redistributed copy stays under LGPL-LR, keeps the licence notice, and says what was changed.
The licence covers the data file only. It puts no conditions on software that just reads the file.
The proper-noun rows, which come from Wikidata, are additionally available under CC0 1.0 with no conditions at all.

The full text is at spdx.org/licenses/LGPLLR.html, and the same notice ships in the README next to the CSV.

Where the data comes from

Two sources, both open:

dict-fr-DELA (pip package, LADL/IGM DELA dictionary) (LGPL-LR)
Wikidata Lexemes (proper nouns) (CC0 1.0)

DELA supplies the common nouns. Wikidata supplies proper nouns such as country and city names, which DELA leaves out. The two have been normalized, deduplicated, and merged, and a spelling that appears in both with conflicting genders is marked epicene. That makes this a modified, derived version of both sources rather than a copy of either.

Using it

The file loads directly into a spreadsheet. For code, for example with pandas:

import pandas as pd

df = pd.read_csv("french-noun-gender.csv")
df[df["word"] == "fromage"]
#         word     gender
#   ...   fromage  masculine

Common questions

Is the French noun gender dataset free to use?

Yes. It is distributed under the LGPL-LR licence, which lets you use and redistribute it, including commercially and in modified form, as long as redistributed copies keep the licence notice and say what was changed. The proper-noun rows from Wikidata are additionally CC0.

What format is it in?

A plain UTF-8 CSV with two columns, "word" and "gender". The gender is masculine, feminine, or epicene. It loads directly into a spreadsheet, pandas, SQLite, or any CSV reader.

What does "epicene" mean in the gender column?

It marks a word that takes either gender, such as a job title used for anyone, or the small number of spellings where the two source dictionaries disagreed. Treat those rows as "either".