MBTI Personality Type Prediction Using BERT-LSTM and Deep Learning on Social Media Posts

It is well known that everyone has a particular ‘personality’ meaning a set of cognitive, affective, and behavioral characteristics. The current study uses text classification based on deep learning models i.e. BERT and LSTM, to predict user's personality traits from their Facebook posts. With the Myers-Briggs type indicator (MBTI) as a baseline, the current research investigates the relationship between user-generated content and personality traits. The results reveal that such an architectural ensemble of BERT contextual embeddings with LSTM's advantages over sequence data indeed results in higher accuracy over personality trait predictions than strategies like Naive Bayes and SVM machine learning algorithms. It was found that personality forecasting systems may benefit from the implementation of advanced NLP models, thus increasing the accuracy and usability of the systems, especially when targeting social media users.