Abstract

Personality assessment is an important topic for researchers and companies, as our personality profoundly influences many of our life choices. In this work, we analyze the task of automated personality assessment by employing only written text. We explore two personality prediction datasets and introduce a new Twitter-based Myers-Briggs personality type dataset, where Twitter users are assigned to one class depending on their self-reported test results. To predict personalities from written text, we tried classification and siamese-based approaches. However, with both approaches, the results are not very promising and are hardly above random guessing for all but one instance. In this case, our models could reliably predict if a person is more into Thinking vs. Feeling and Judging vs. Perceiving.

Sources

Thumbnail: The Big Five personality traits