A Myers-Briggs personality predictor based on user social media posts [Facebook / Twitter supported].
See working demo at https://aiae.tk/crystal
Data is sourced from https://www.kaggle.com/datasnaek/mbti-type
Prediction pipeline:
- Initial cleaning via tweet-preprocessor [https://github.com/s/preprocessor]
- Word embeddings via GloVe trained on tweet data [https://nlp.stanford.edu/projects/glove/]
- Sentence embeddings via Bag-of-words
- XGBoost [similar to forest of random decision trees] for predictions
Backend:
- Flask
Frontend:
- React
- Redux
Clone git repository. Create a Python3 virtual environment via venv or Anaconda. Install requirements [pip install -r requirements.txt]. Run Flask server within Flask_Server directory [python app.py]