name-gender-nn-rs

Determines whether a first name is male or female. Ships with two datasets:

• Russian (data/names_ru.json) — Cyrillic alphabet, 108 names
• English (data/names_en.json) — Latin alphabet, 100 names

Each dataset defines its own alphabet. You train a separate model per dataset, and each model works only with names from its own alphabet.

Same as the Python reference:

• Alphabet. Each dataset defines its own alphabet as a string of characters plus a _ pad character. The Russian dataset uses the 33-letter Cyrillic alphabet; the English dataset uses the 26-letter Latin alphabet.
• Features. Each name is split into three blocks: letters from the start, letters from the end, and the 2-letter suffix. Every character is one-hot encoded over the dataset alphabet.
• Model. Fully connected: input → hidden layer (40 neurons, sigmoid) → output (1 neuron, sigmoid). Trained with SGD, lr = 0.1, 8000 epochs.
• Data. Russian: 108 names (49 female, 59 male). English: 100 names (50 female, 50 male).

git clone https://github.com/smartlegionlab/name-gender-nn-rs.git
cd name-gender-nn-rs
cargo build --release

Binary: target/release/name-gender-nn-rs.

Requirements: Rust 1.85+ (2024 edition), Cargo.

All commands are subcommands of a single binary.

Validate a dataset

cargo run --release -- check --data data/names_ru.json
cargo run --release -- check --data data/names_en.json

Output for the Russian dataset:

File: data/names_ru.json
Alphabet size: 34
Loaded 49 female and 59 male names

Alphabet check: OK
Duplicate check: OK
Cross-gender check: OK

All checks passed.

Checks: all names use only characters from the dataset alphabet; no duplicates inside each list; no name appears in both female and male.

Train

cargo run --release -- train --data data/names_ru.json --out weights_ru.json
cargo run --release -- train --data data/names_en.json --out weights_en.json

Output for the Russian dataset:

Loaded 108 names from data/names_ru.json
Alphabet size: 34
epoch     0  error=13.671195  t=0.0s
epoch   500  error=0.019290  t=0.1s
...
epoch  7500  error=0.000797  t=0.9s

Train accuracy: 108/108 = 100.0%
Weights saved to weights_ru.json
Total time: 1.0s

Predict

cargo run --release -- predict --weights weights_ru.json александр
cargo run --release -- predict --weights weights_en.json alexander

One-liner output:

александр -> male  (confidence 99.7%)
alexander -> male  (confidence 99.8%)

Interactive mode (no name argument):

cargo run --release -- predict --weights weights_ru.json

Enter a name (or 'exit' to quit):
> анна
анна -> female  (confidence 99.8%)
> exit

Measured on the same machine as the Python reference implementation.

Same algorithm, same accuracy. The speedup comes from the release build profile (opt-level = 3, lto = true, codegen-units = 1).

Inputs from the datasets

анна       -> female  (confidence 99.8%)
дмитрий    -> male    (confidence 100.0%)
ольга      -> female  (confidence 99.5%)
александр  -> male    (confidence 99.7%)
екатерина  -> female  (confidence 100.0%)
mary       -> female  (confidence 99.0%)
john       -> male    (confidence 100.0%)
elizabeth  -> female  (confidence 99.7%)
alexander  -> male    (confidence 99.8%)
catherine  -> female  (confidence 99.9%)

Full forms not in the datasets

сара       -> female  (confidence 85.5%)
мара       -> female  (confidence 93.9%)
федот      -> male    (confidence 96.1%)
жанна      -> female  (confidence 99.9%)
ева        -> female  (confidence 62.1%)
karl       -> male    (confidence 95.0%)
victoria   -> female  (confidence 94.9%)
diana      -> female  (confidence 99.7%)
george     -> male    (confidence 100.0%)
peter      -> female  (confidence 99.7%)   wrong

19 out of 20 correct. The single failure (peter) is a known limitation of the small training set: the model has never seen a short male name ending in -er, and the closest patterns in the training data are female (elizabeth, catherine, mary).

name-gender-nn-rs/
├── Cargo.toml
├── Cargo.lock
├── data/
│   ├── names_ru.json   # Russian dataset (Cyrillic)
│   └── names_en.json   # English dataset (Latin)
├── src/
│   ├── main.rs         # CLI entry point (clap)
│   ├── model.rs        # NameNet: forward, train_step, save, load
│   ├── train.rs        # training
│   ├── predict.rs      # inference
│   └── check_data.rs   # dataset validation
├── LICENSE             # BSD 3-Clause License
├── DISCLAIMER.md       # full legal disclaimer
└── README.md

weights_*.json files are generated by train and are not tracked by git.

Open the dataset file and add a name to either female or male:

{
  "alphabet": "абвгдеёжзийклмнопрстуфхцчшщъыьэюя_",
  "female": ["анна", "...", "александра"],
  "male":   ["иван", "...", "александр"]
}

Then retrain:

cargo run --release -- train --data data/names_ru.json --out weights_ru.json

Create data/names_XX.json with the correct alphabet and name lists, then train:

cargo run --release -- train --data data/names_XX.json --out weights_XX.json
cargo run --release -- predict --weights weights_XX.json

The model reads the alphabet from the dataset, so any language with a single-alphabet script works out of the box.

• name-gender-nn-py — Python reference implementation. Same algorithm, same datasets, intended as an educational project.

By using this software, you agree to the full disclaimer terms.

Software provided "AS IS" without warranty. You assume all risks.

Full legal disclaimer: See DISCLAIMER.md

License: BSD 3-Clause License