Identifying personality and topics of social media

Abstract

Twitter and Facebook are the renowned social networking platforms where users post, share, interact and express to the world, their interests, personality, and behavioral information. User-created content on social media can be a source of truth, which is suitable to be consumed for the personality identification of social media users. Personality assessment using the Big 5 personality factor model benefits organizations in identifying potential professionals, future leaders, best-fit candidates for the role, and build effective teams. Also, the Big 5 personality factors help to understand depression symptoms among aged people in primary care. We had hypothesized that understanding the user personality of the social network would have significant benefits for topic modeling of different areas like news, towards understanding community interests, and topics. In this thesis, we will present a multi-label personality classification of the social media data and topic feature classification model based on the Big 5 model. We have built the Big 5 personality classification model using a Twitter dataset that has defined openness, conscientiousness, extraversion, agreeableness, and neuroticism. In this thesis, we (1) conduct personality detection using the Big 5 model, (2) extract the topics from Facebook and Twitter data based on each personality, (3) analyze the top essential topics, and (4) find the relation between topics and personalities. The personality would be useful to identify what kind of personality, which topics usually talk about in social media. Multi-label classification is done using Multinomial Naïve Bayes, Logistic Regression, Linear SVC. Topic Modeling is done based on LDA and KATE. Experimental results with Twitter and Facebook data demonstrate that the proposed model has achieved promising results.