Sentiment Valences for Automatic Personality Detection of Online Social Networks Users using Three Factor Model

The interactions using text or writing has become an important medium to communicate among human. The inventions of online networks technologies and applications such as social networks rapidly growth the size of digital textual data and indirectly raises the curiosities to mine the juicy information encapsulated in the text data. Researchers have stated that linguistics of the people either in the online or offline having strong correlations with their self-disclosure. The mass creations of digital textual data raised the intentions of scholars to study the hidden abstractions of personality by applying automatic personality detection approaches. Most of the current automatic personality detection studies focused on Big 5 personality model as a framework to study the underlying characteristics of human. As such, this study incorporates the Three Factor Personality (PEN) Model as a personality framework to guide our understanding and revealing the role of words in depicting the characteristics of a user. This preliminary study revealed how the sentiment perceptions of public towards specific words could assist us in detecting the personality of Facebook users by exploiting their status messages. As the first phase of our study, this experiment focuses on gathering the general perceptions of Malaysians towards 52 English words and 17 interjections that was retrieve from the domain stated above. A Likert scale questionnaire executed to find the sentiment valences of 67 words through analyzing responses from participants and statistical significance will assist the categorization of words under PEN traits. The evaluation provides the necessary analysis that could assist our main research that focuses on automatic personality detections specifically on the psychoticism trait. Our initial findings has highlighted that the five words categorized under psychoticism has strong Cronbach's Alpha coefficients and significant effect from multivariate analysis that indirectly affirmed the reliability of the categorization of the words. (C) 2015 The Authors. Published by Elsevier B.V.