A Deep Learning Approach to Personality Identification

Identifying personality traits is a crucial aspect of mental healthcare, which may be effectively achieved through integrating deep learning techniques within the field of psychology. Particularly, “personality” distinguishes an individual from others by describing feelings, thinking, and behaviors. Determining the correct personality proposes a good understanding of the individual and supports the psychologist in diagnosing and treating patients. In this paper, we developed a deep learning approach to identify personality from the MBTI (Myers-Briggs Type Indicator) Kaggle dataset. The model included three main components: convolutional layers to extract information, the LSTM layer that processes information in long-term dependency, and fully connected layers to classify the personality. As a result, our model outperformed the existing benchmarks of several studies in the 16-personalities classification task. Our models also led in performance in 2 over 4 binary personality dimension classification tasks. With these investigations, we recognized that our models could be applied to classify personality with better accuracy than several existing models.