Modeling of User Portrait Through Social Media

Haiqian Gu, Jie Wang, Ziwen Wang, Bojin Zhuang, Fei Su

Abstract

Nowadays, massive useful data of user information and social behavior have been accumulated on the Internet, providing a possibility of profiling user's personality traits online. In this paper, we propose a psychological modeling method based on computational linguistic features to profile Big Five personality traits of users on Sina Weibo (a Twitter-like microblogging service in China) and their correlations with user's social behaviors. To the best of our knowledge, this is the first research on investigating the potential relationship between profile information, social-network behaviors and personality traits of users on Sina Weibo. Our results demonstrate an effective modeling approach to understanding demographic and psychological portraits of users on social media without customer disruption, which is useful for commercial incorporations to provide better personalized products and services.

Introduction

Traditional personality assessment has long relied on self-reported questionnaires, which, while accurate, are often expensive, time-consuming, and difficult to scale for large populations. With the explosion of social media, platforms like Sina Weibo have become rich repositories of natural digital footprints. This paper addresses the challenge of "user portrait" modeling by integrating computational linguistics, social behavior, and demographic data to predict the "Big Five" personality traits (Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism) non-intrusively.

Methodology

The researchers developed a multi-stage pipeline to bridge the gap between raw social data and psychological traits:

  • Linguistic Analysis: The study utilized SC-LIWC (Simplified Chinese version of Linguistic Inquiry and Word Count), a dictionary-based tool that categorizes words into psychologically meaningful groups such as "Affect," "Social," and "Cognitive processes."
  • Mapping Matrix: To calibrate the model, ground-truth Big Five scores were collected from 100 volunteers. A mapping matrix was then calculated between these volunteers' SC-LIWC word frequencies and their actual personality scores using the Scikit-learn library.
  • NLP Preprocessing: For the broader dataset, the Jieba segmentation tool was used for Chinese text processing. Rigorous cleaning was implemented to remove noise, including URLs, system-generated messages, and spam advertisements.
  • Correlation Analysis: Pearson correlation analysis was performed to identify significant relationships between personality scores and various features like emoticon usage, tag diversity, and profile metadata.

Psychological Markers in Social Media

The study reveals distinct correlations between digital behavior and personality dimensions. For instance, Extraverts use first-person pronouns (e.g., "I," "myself") more frequently, while Agreeable individuals favor collective terms like "We."

Figure 1: Word clouds showing top related tags for Openness, Conscientiousness, Extraversion, and Agreeableness. High Openness users favor "Reading" and "Campus Life," while Low Openness users are more interested in "Zodiac" and "Anime".

One of the most striking findings is the "Sleep" signal: users who tag themselves with "Sleep" tend to score high in Neuroticism but low in Agreeableness and Extraversion. Furthermore, high Neuroticism is strongly correlated with "Ingest" words (eating, drinking), suggesting a link between emotional instability and mentions of physical consumption.

Figure 2: Word clouds for High and Low Neuroticism, illustrating the prevalence of tags like "Sleep," "Living," and "Weirdo" among high-scoring individuals.

The model successfully identified statistically significant markers for all five personality dimensions across a dataset of 6,467 valid users:

  • Demographics: In this sample, men were generally more conscientious, while women scored higher in extraversion and neuroticism.
  • Age Trends: The data suggests a psychological "maturing" process; as users age (within the 10-47 range), they tend to score higher in Conscientiousness and Agreeableness, while Openness and Extraversion decline.
  • Digital Self-Disclosure: A positive correlation exists between the length of a user's self-introduction and their scores for Conscientiousness and Extraversion, indicating that more detailed bios signal a higher willingness for social engagement.
  • Verification Status: Verified users (influencers or professionals) typically scored higher in Conscientiousness compared to unverified users.

Conclusion

This research demonstrates that social media behavior can serve as a reliable proxy for psychological profiling. While the current model is limited by the initial calibration sample size and the presence of fake demographic data, it paves the way for "precision marketing" and personalized services. Future work will focus on synthesizing data across multiple platforms, such as Netease Music and Taobao, to link personality profiles with specific consumer habits and lifestyle preferences.

Summary updated

Abstract

Nowadays, massive useful data of user information and social behavior have been accumulated on the Internet, providing a possibility of profiling user's personality traits online. In this paper, we propose a psychological modeling method based on computational linguistic features to profile Big Five personality traits of users on Sina Weibo (a Twitter-like microblogging service in China) and their correlations with user's social behaviors. To the best of our knowledge, this is the first research on investigating the potential relationship between profile information, social-network behaviors and personality traits of users on Sina Weibo. Our results demonstrate an effective modeling approach to understanding demographic and psychological portraits of users on social media without customer disruption, which is useful for commercial incorporations to provide better personalized products and services.
BibTeX
@inproceedings{gu2018modeling,
  author = {Haiqian Gu and Jie Wang and Ziwen Wang and Bojin Zhuang and Fei Su},
  title = {{Modeling of User Portrait Through Social Media}},
  booktitle = {ICME},
  year = {2018},
  url = {https://openreview.net/forum?id=bAWOVrVWrg},
}
Figure