Personality Detection on Social Media Twitter Using Long Short-Term Memory with Word2Vec

Abstract

Personality is one of the things that exist in every human being. From this personality, it will be seen that humans are unique and diverse individuals. One of the causes is because each human being has a personality. This personality will create humans to have a way of thinking and behave. Personality can also be a benchmark or valuation. For example, in employee recruitment, some companies have standards for candidate workers to be recruited. One tool for doing this is a psychological test. But the drawback of this psychological test method is that it takes a long time. So we created a system that can predict personality based on the user's tweet data on Twitter because it's contain many opinions or thoughts from the user that can describe the user's personality. The methods we use are Word2Vec for word embeddings and LSTM (Long Short-Term Memory) for predicting. The system created focuses on seeing how a collection of tweets can predict a user's personality. The first scenario it's to combine every two rows of tweet and have a scoring accuracy 45%. The second scenario it's to combine every ten rows of a tweet and have the best accuracy of 48%. The system uses five labels of personality that we call the Big Five Personality.