This project is an application of a names classifier, that is, classify a name of either a girl's name of boy's name. It is built as an ilustration of building machine learning applications, but the performance of the classifier itself is of no interest.
Up to now, this only works with spanish names.
I expect that the particular choice for problem and solution approach is of no real value (from the machine learning perspective), but let me know if you can think of some ;)
The slides for the tutorial can be downloaded here
First of all I recommend you install the requirements in a virtual environment (see instructions for pip and virtualenv) The application runs by issuing the following command in the terminal: ` python local_app.py --log The application has 4 relevant commands:
train: to train the model on a given dataset.test: to test the model on a given dataset.infer: to infer whether a given name is that of a girl or a boy. Use a quoted name in case of names with spaces on it.download-data: to download the dataset of names, parse, split the data into train and test sets and store it as .hdf5 file.
Please use the --help option for more details.
-
First make sure you intall the additional requirements: Zappa and boto3
-
Deploy the app in your computer runing
pyhton microservice.pyon the command line or in AWS Lambda using the Zappa commandzappa deploy dev. -
Use the app by making an http GET request to
BASE_URL/infer/<name>and you will get the corresponding label (Man or Woman). Note thatBASE_URLdepends on where it is deployed.