Article Preview
Top1. Introduction
Personality is one of the typical and enduring topics in psychology. Personality prediction can be basically defined as identifying personality traits of a person by using a set of data. Current personality studies often rely on data from well-controlled and specified environments (Rozin, 2001), such as survey studies. However, social network usage is getting more and more popular and the data produced by social media users can provide a valuable resource to automatically determine human personality. Social media is one of the most commonly used services on the Internet. Studies show that one third of the time spent on Internet is allocated on social media sites (Suhartono et al., 2017). Although the use of social networks is increasing day by day, the number of studies focusing on the relation of social media and personality is still limited, and there are open problems to be studied such as analyzing the effective and language dependent features and the use of deep neural models for the problem (Ahmad, 2020) (Bharadwaj, 2018).
Psychologists show that there is a strong relationship between the personality and the language behaviour of individuals (Fast & Funder, 2008). Nowadays, computer based written communications became as common as face to face communication. Therefore, there are vast number of research efforts to determine personality traits of individuals form the texts they have written. Among the mediums, social media is probably by far the most popular venue people use. However, people write very informally and short texts in those places. Therefore, in this work we focus on such kind of texts and aim to determine the personality traits of individuals from their short informal texts.
Most of the previous studies about personality analysis are conducted on samples provided under controlled conditions, such as talks or texts on given topic (Baddeley & Singer, 2009), (Fast & Funder, 2008), (Hirsh & Peterson, 2009), (Pennebaker & King, 2000). On the other hand, naturalistic approach is proven to be more powerful in (Mehl, Gosling, & Pennebaker, 2006), where samples of participants’ natural language use and behavior are recorded and analyzed. The study in (Mehl, Gosling, & Pennebaker, 2006) provides useful results about the relation between natural language use and personality, most of which were not captured in laboratory studies.
Identifying users’ personality can be useful in different domains such as customer analysis for commerce, social psychology or recommendation systems to build personalized models and to improve user experiences. Advertising can be another field that can benefit from personality analysis. In (Odekerken, De Wulf, & Schumacher, 2003), the relation between consumer personality and marketing techniques is presented. In the last decades, psychologists who work on personality analysis by using lexical approach, ended up viewing personality in five dimensions (Boele, 2000). This approach is known as Big Five personality traits (Goldberg, 1990). This model alleges that dimensions of neuroticism, openness, extraversion, agreeableness and conscientiousness can comprise most of the structure of the personality traits. Big Five personality traits are known as the most widely accepted personality dimensions in psychology (Zhang, 2002).