每周论坛之十（总第225期）：龚涛（美国教育考试服务中心）,2019年11月11日下午
每周论坛之十（总第225期）
报告人：龚涛 博士
报告人单位：Educational Testing Service
题目：Studying psychological traits using data social media language
时间：2019年 11月11 日 （周一，14:20-16:00）
地点：中山大学东校园心理学系（南学院楼C座）305
报告简介：
The large amount of language data accumulated in popular social media websites could reflect social relations, personal emotions, and life events of social media users, thus qualified as important sources of public language corpora for academic research, public health, and policy making. Advancement in natural language processing and data science also makes social media based psychological studies a hot interdisciplinary topic in personality and social psychology. This talk presents two example studies using social media language to study individual psychological traits.
The first study proposed a unified approach to construct a profile of subjective well-being based on Facebook status updates. We applied sentiment analysis to generate users affect scores, and trained a random forest model to predict SWL using affect scores and other language features of the status updates. Analyses showed that: the computer-selected features resembled the key predictors of SWL as identified in early studies; the machine-predicted SWL was moderately correlated with the self-reported SWL based on the Center for Epidemiological Studies Depression Scale (CES-D), indicating that language-based assessment could constitute valid SWL measures; and the machine-predicted subjective well-being profile could also reflect other psychological traits like depression.
Affect describes a personal feeling or emotion in reaction to stimuli. In social media, affective expressions are found to be related to mental illness such as depression. The second study makes use of self-reported CES-D surveys and three years worth of Weibo posts preceding the surveys of 1,664 Chinese Weibo users to show that: increase of positive and negative affects in social media posts of users with moderately high depressive symptoms occurred years before they completed the surveys; the presence of stressful life events dramatically increased their risks of developing high depressive symptoms within three months after the occurrence of events; and rumination is an important component of contents of negative affect and positive affect is often found in rumination focusing on coping strategies.
These studies are insightful for developing automatic monitor tools of psychological traits using social media language, and expand our understanding of the longitudinal affect patterns on social media, correlation between stressful life events and individuals risks of developing depressive symptom. They are also useful for identifying vulnerable individuals to depression or other mental illness and (re)allocating clinical resources for sufficient and timely support.
报告人简介：
Dr. Tao Gong graduated in 2007 at the Chinese University of Hong Kong, supervised by Prof. William S-Y. Wang, a globally famous linguist. After graduation, he had taken the postdoctoral job at Department of Physics, Sapienza University of Rome, the Humboldt Fellowship at Max Planck Institute for Evolutionary Anthropology in Germany, and the British Academy of Sciences Visiting Fellowship at University of Edinburgh in UK. After teaching for four years as a research assistant professor in Department of Linguistics, University of Hong Kong, Dr. Gong moved to the US in 2014, and worked as a research scientist at Haskins Laboratories, Yale University. In 2017, he joined Educational Testing Service as a senior research scientist (data science).
His current research interests include: evolutionary linguistics, computational simulation, complex systems, psychology, data science and machine learning.
His publications include a monomgraph Computational Simulation in Evolutionary Linguistics: A Study on Language Emergence (Frontiers in Linguistics Monograph IV, published in 2009 by Institute of Linguistics, Academia Sinica, Taipei), and over 60 journal papers in many interdisciplinary and linguistics journals, such as Scientific Reports, PNAS, Proceedings of Royal Society B, Frontiers in Psychology, Language, Lingua, Language Sciences, Artificial Life, Biolinguistics, PLOS ONE, and so on. He also gave several reports in many international conferences such as Evolang 5-9, Protolang 1&6, IACL 19, and CIEL 10&11.
