This project predicts the gender of a sentence that some users write. Given any sentence, the model will output male or female which represents the gender of the user who writes that sentence. We deployed a Django website for user interaction, displaying the result and related evidence in real time. We used L1/L2 norm to extract the most significant words/n-grams that determine the prediction. We also used the package Lime Text Explainer to further support the prediction based on the confidence interval of each feature.
To run the website, first install Django and LIME.
pip install Django pip install lime
Refer to the following sites for full installation instructions. https://docs.djangoproject.com/en/2.2/topics/install/ https://github.com/marcotcr/lime
After the packages are install, inside nLpDecipher/decipher, run
python manage.py runserver
Then use a browser to open 127.0.0.1:8000
After opening the website, click the "Train Model" button to train models.
After the "Training Done" text appears below the button, we can now input text in either of the textbox and then click the button below to predict.
The predictions should show up shortly after one of the buttons is clicked.