User-level psychological stress detection from social media using deep neural network

Abstract

It is of significant importance to detect and manage stress before it turns into severe problems. However, existing stress detection methods usually rely on psychological scales or physiological devices, making the detection complicated and costly. In this paper, we explore to automatically detect individuals' psychological stress via social media. Employing real online micro-blog data, we first investigate the correlations between users' stress and their tweeting content, social engagement and behavior patterns. Then we define two types of stress-related attributes: 1) low-level content attributes from a single tweet, including text, images and social interactions; 2) user-scope statistical attributes through their weekly micro-blog postings, leveraging information of tweeting time, tweeting types and linguistic styles. To combine content attributes with statistical attributes, we further design a convolutional neural network (CNN) with cross autoencoders to generate user-scope content attributes from low-level content attributes. Finally, we propose a deep neural network (DNN) model to incorporate the two types of userscope attributes to detect users' psychological stress. We test the trained model on four different datasets from major micro-blog platforms including Sina Weibo, Tencent Weibo and Twitter. Experimental results show that the proposed model is effective and efficient on detecting psychological stress from micro-blog data. We believe our model would be useful in developing stress detection tools for mental health agencies and individuals.

Researching Mental Health Disorders in the Era of Social Media: Systematic Review

“…It is employed to classify the polarity of a given text into categories such as positive, negative, and neutral [77]. Several studies [24,28,30,32,34,39, 49,50,52-55,57,60,65-68,70] used the well-known linguistic inquiry and word count (LIWC) [78] to extract potential signals of mental problems from textual content (eg, the word frequency of the first personal pronoun “I” or “me” or of the second personal pronoun, positive and negative emotions being used by a user or in a post). OpinionFinder [79] was used by Bollen et al [71] and SentiStrength [80] was used by Kang et al [27] and by Durahim and Coşkun [47] to carry out sentiment analysis.…”

“…Some studies investigated these images for cues of mental disorders [27,57]. Color compositions and scale-variant feature transform descriptor techniques were used to extract emotional meanings of each individual image [27].…”

“…Color compositions and scale-variant feature transform descriptor techniques were used to extract emotional meanings of each individual image [27]. Image properties, comprising color theme, saturation, brightness, color temperature, and color clarity, were analyzed be Lin et al [57]. …”

Prospects and challenges for social media data in conservation science

“…Going through social media contents manually is, however, a laborious task. Luckily the tools for automatic data classification that may help in simple content analyses are developing quickly (e.g., Deng et al, 2009;Laurance et al, 2009;Angus et al, 2010;Ozel and Park, 2012;Lin et al, 2014;Zhou et al, 2014;Karpathy and Fei-Fei, 2015;Russakovsky et al, 2015;Schwartz and Ungar, 2015). Practical applications utilizing such automatic classification approach are also emerging by using e.g., image recognition, such as the smartphone app CamFind (http://camfindapp.com/).…”

Detection of Stress and Relaxation Magnitudes for Tweets

“…It extracts the tweet-level linguistic, visual and social attributes defined in [20]. These attributes are fed into cross auto-encoders which are embedded in a CNN, integrating them to user level content attributes [21]. This system recommends links to users to dissipate stress.…”