The disproportionate often irresponsive use of Machine learning Algorithms risks to emphasise stereotypes already present in data. Whenever data reflects biases of the broader society, the learning algorithm captures and learns from these stereotypes.
On this website we visualise how text embeddings trained on Google News represent gender.
Install PIP requirements with pip install -r requirements.txt
.
Remember to insert a .txt
or .bin
embedding file in gender.py
, you can create one here: https://github.com/sinanatra/create-text-embedding
or fetch it from here: https://nlp.stanford.edu/projects/glove/
And to download spacy modules from here: https://spacy.io/usage/models
Run FLASK_APP=server.py flask run
to start.
Run FLASK_APP=server.py FLASK_DEBUG=1 flask run
to start in debugging mode.