theryston / gender-name

An API that uses artificial intelligence to predict someone's gender based on their name

Gender Name: AI-Powered Gender Prediction

Gender Name is a powerful and sophisticated tool that employs state-of-the-art artificial intelligence techniques to accurately predict a person's gender based on their name. This versatile tool has numerous applications across various industries, including demographic analysis, marketing personalization, user experience enhancement, and social science research.

Core Features

• Dedicated models for different cultures and languages.
• User-friendly API endpoint designed for seamless integration with a wide range of platforms and applications.
• Continuously updated and refined models to maintain high performance and accuracy.
• Totally free, open source and without request limits

Examples

• Predict USA gender based on John Doe name
• Predict BR gender based on Theryston name

Available Models

Model	Country	Description
gnbr	Brazil	Predict gender by names of humans from Brazil
gnusa	USA	Predict gender by names of humans from USA

Multilingual and Multicultural Support

Gender Name is designed to support names across various cultures and languages. We have dedicated models for different cultures and languages, each optimized for the naming conventions and gender associations of a particular country or culture. This provides high accuracy and cross-cultural applicability.

Seamless Integration with Diverse Platforms

The Gender Name API has been designed to ensure effortless integration with a variety of platforms and programming languages. By making a GET request to the provided API endpoint and submitting the desired name, you will receive a JSON response containing the predicted gender.

Community Contributions and Collaboration

We actively encourage and appreciate contributions from the community to further enhance the accuracy and functionality of Gender Name. If you have any suggestions, ideas for improvements, or model refinements, please feel free to create an issue or submit a pull request on our repository. Together, we can make this tool even more powerful and effective, serving a wide range of languages and cultures.

Models

You can list all available models by accessing the GET /models endpoint:

GET https://api.gendername.org/models

Detailed API Usage Instructions

To use the Gender Name API it is very simple, just make a request to GET /predict passing the name and model in the query params, below is the more detailed request:

Endpoint:

GET https://api.gendername.org/predict?name={person_name}&model={model_name}

• Replace {person_name} with the name of the person you want to predict (can be up to 1000 names if you split it by comma, like: {person_name1},{person_name2}).
• Replace {model_name} with the model you wish to use for the prediction.

Response Example:

{
  "elapsed_ms": 2.2819042205810547,
  "model_name": "gnbr",
  "results": [
    {
      "gender": "male",
      "name": "joão",
      "probability": 0.9888328313827515,
      "probabilities": {
        "female": 0.0000666277264826931,
        "male": 0.9888328313827515,
        "unisex": 0.011100534349679947
      },
    }
  ]
}

License

The Gender Name API are licensed under the MIT license.

gender-name's Issues

Configure Prisma and DB

We need to have a database to store the models, users, API key...

To do this, we have to configure Prisma with a database URL. The database will run in vercel while the API request is not much

Some names, gender seems to be incorrect

Please, check these names, the gender seems to be incorrect.
DOMINIQUE, HELLEN, JAZA, JOYCE, JANE, ESTER, GEOVANE, GIOVANNE, IASMYN, JOYCE, JHENIFER, KEVIN, LUZITANIA, JOYCE, NEUZIMAR, NEUZIMAR, SHIRLEI, ROSECLER, THAIS, THAIS, THAIS, THAIRONE, TAIS.
Thanks.

If neither the "model" nor the "country" parameter is provided, the API should default to using the GNB-2 model.

The API should allow an optional parameter called "model", through which users can specify the model they wish to use.

Integrate API with new GNB-2 model

This issue addresses some significant changes in the issue #1, including the introduction of a new model called GNB-2. The main goal of this issue is to integrate the new GNB-2 model into the existing API.

Tasks to be accomplished:

• #3
• #4
• #5

Limits of consults not specified and result in error

Error after ~3 hundred requests:
{'statusCode': 500, 'message': 'Internal server error', 'timestamp': 'AAAA-MM-DDT00:MM:SS.MMMZ', 'path': '/predict?model=gnbr&name=name'}
IP locked or banned? I suggest instructing about the limits of requests.

The API should accept an optional parameter called "country". If provided, the API should search for and use the country-specific model.

Train new model GNB-2

After receiving various comments on the Tabnews introductory article, we identified issues with our artificial intelligence model. We have decided to develop a new model called Gender Name Brazil 2 (GNB-2) to address these concerns.

Main issues identified:

1. The API does not take the geographical context into account when identifying a person's name (e.g., "Yuri" can be a female name in Japan and a male name in Russia).
2. In some cases, the model may incorrectly predict the gender.
3. For unisex names, the AI returns a random gender.

Proposed solutions:

1. We will train multiple models for each country, and the API will select the appropriate model based on the parameters provided in the URL.
2. The current model (Gender Name) was trained with only 3,000 Brazilian names. The next model will be trained with a larger number of names to improve its accuracy.
3. We will add a new category called "unisex" to distinguish names that can be used by both genders.

Add a demo to home page

The demo should have a input for the name and a select the model and a button to run the models