Applied Intelligence
https://doi.org/10.1007/s10489-018-1212-4
Deep learning-based personality recognition from text posts of online
social networks
Di Xue
1
· Lifa Wu
1
· Zheng Hong
1
· Shize Guo
2
· Liang Gao
2
· Zhiyong Wu
1
· Xiaofeng Zhong
3
· Jianshan Sun
4
© Springer Science+Business Media, LLC, part of Springer Nature 2018
Abstract
Personality is an important psychological construct accounting for individual differences in people. Computational
personality recognition from online social networks is gaining increased research attention in recent years. However, the
majority of existing methodologies mainly focused on human-designed shallow statistical features and didn’t make full use
of the rich semantic information in user-generated texts, while those texts are exactly the most direct way for people to
translate their internal thoughts and emotions into a form that others can understand. This paper proposes a deep learning-
based approach for personality recognition from text posts of online social network users. We first utilize a hierarchical
deep neural network composed of our newly designed AttRCNN structure and a variant of the Inception structure to learn
the deep semantic features of each user’s text posts. Then we concatenate the deep semantic features with the statistical
linguistic features obtained directly from the text posts, and feed them into traditional regression algorithms to predict the
real-valued Big Five personality scores. Experimental results show that the deep semantic feature vectors learned from our
proposed neural network are more effective than the other four kinds of non-trivial baseline features; the approach that
utilizes the concatenation of our deep semantic features and the statistical linguistic features as the input of the gradient
boosting regression algorithm achieves the lowest average prediction error among all the approaches tested by us.
Keywords Personality recognition · Deep learning · Online social networks · Big Five personality
1 Introduction
Personality is a psychological construct aimed at explaining
various human behaviors in terms of a few stable and
measurable individual characteristics [1]. It not only reflects
an individual’s consistent patterns of behavior, thought
and interpersonal communication [2], but also influences
important life aspects [3], including happiness, motivations,
preferences, emotion, mental and physical health [4]. The
dominant paradigm for formal description of personality
in psychology [1] is the Big Five, also known as the Five
Factor Model, which consists of five basic traits: openness
Zheng Hong
hz5215@163.com
1
Army Engineering University, Nanjing 210007, China
2
Institute of North Electronic Equipment, Beijing 100083,
China
3
Electronic Engineering Institute, Hefei 230037, China
4
Hefei University of Technology, Hefei 230009, China
to experience (O), conscientiousness (C), extraversion (E),
agreeableness (A), and neuroticism (N) [5]. The study of
personality is foundational to psychology, and personality
recognition (PR) [6] can benefit many other applications,
such as social network analysis [7], recommendation
systems [8], deception detection [9], authorship attribution
[10], sentiment analysis/opinion mining [11] and so on [12].
However, the traditional methods of personality assessment
through questionnaire investigation or expert interview are
costly and less practical in cyber space [13].
Along with the popularization of social media in recent
years, automatic personality recognition from online social
network (OSN) is gaining increased research attention
because of its potentials in many computational applica-
tions [14]. There are rich self-disclosed personal informa-
tion and emotional contents on social media, which have
been proved to be highly correlated with user’s person-
ality traits [15–22]. Numerous studies have been done to
explore optimal feature space and machine learning algo-
rithms for recognizing individual’s personality [23–28].
However, the achievements of existing methodologies are
not satisfactory. For one thing, the majority of existing
@Phil_Robichaud
@deepthiw
@JoseServera