Model for Personality Detection Based on Text Analysis

Abstract

Personality is a unique trait which distinguish people from each other. It is a set of individual differences in thinking, feeling and behaving of people, and it affects interaction, relationships and environment of people. Personality can be useful to several tasks like education, training, marketing and personnel recruitment. Several methods to detect personality have been proposed and there are several psychological models proposing different personality dimensions. Previous research states that personality can be detected by means of text analysis. We have built a model for personality detection based on statistical analysis of language and DISC model. As fundamental components of the model, we built a linguistic corpus with personality annotations and a corpus of words related to personality. To build the model, we conducted a study where 120 individuals participated. The study consisted in filling a personality test and writing some paragraphs. We trained several machine learning algorithms with data from the study, and we found Sequential Minimal Optimization algorithm achieved best results in classification.