Social RoleAware Emotion Contagion in Image Social Networks

  • Slides: 29
Download presentation
Social Role-Aware Emotion Contagion in Image Social Networks Yang, Jia, Boya Wu, and Jie

Social Role-Aware Emotion Contagion in Image Social Networks Yang, Jia, Boya Wu, and Jie Tang Department of Computer Science and Technology Tsinghua University

Image social network (e. g. , Flickr) users post photos, which express their emotional

Image social network (e. g. , Flickr) users post photos, which express their emotional statuses. 2

Users are connected … st i x e n o i g a t

Users are connected … st i x e n o i g a t n o c n io t o m E s? s k r o Doe w t e n l a i c o s in image Emotion Contagion: The cascade of users’ emotional statuses influence each other 3

Social Roles of Users Bad boys League of heroes Opinion leaders: users taking central

Social Roles of Users Bad boys League of heroes Opinion leaders: users taking central positions in communities 4

Social Roles of Users Bad boys League of heroes e c n e u

Social Roles of Users Bad boys League of heroes e c n e u l f n i s e l o r l a i Will soc s? n o i g a t n o c n o i t emo Structural hole spanners: users bridge otherwise disconnected communities 5

Predicting Users’ Emotional Status • Input: An image social network G=<V, M, E, R>,

Predicting Users’ Emotional Status • Input: An image social network G=<V, M, E, R>, where V is a set of users, M is a set of images, E represents following relationships between users, and each element in R (v, m, t) denotes that user v publishes image m at time t. • We use a matrix Y to denote users’ emotional status, where yvt indicates v’s emotion at time t. yvt {happiness, surprise, anger, disgust, fear, sadness} • Task: Given G, Y, a time stamp t, our goal is to learn [1] Yang, Y. ; Jia, J. ; Zhang, S. ; Wu, B. ; Chen, Q. ; Li, J. ; Xing, C. ; and Tang, J. 2014. How do your 6 friends on social media disclose your emotions? In AAAI’ 14, 2014.

Related Work Predicting users’ emotions by jointly modeling images and comments. Yang, Y. ;

Related Work Predicting users’ emotions by jointly modeling images and comments. Yang, Y. ; Jia, J. ; Zhang, S. ; Wu, B. ; Chen, Q. ; Li, J. ; Xing, C. ; and Tang, J. 2014. How do your friends on social media disclose your emotions? In AAAI’ 14, 2014. ’ s r e s u t cconsidering i d e r p r Predicting users’ emotions in mobile network by te t e b o t g w n i r Hologs. e calling/messaging d i ns o c y b s Tang, J. ; Zhang, Y. ; Sun, J. ; Rao, A. Quantitative study t. J. ; io. Yu, n. W. ; Chen, Y. ; and Fong, n o m e ? s o i of individual emotional states in social networks. TAC’ 12, 2012. g a nt o c n o i t emo Images drive event engagement (e. g. , clicking “like” or adding comments) 100 times faster than text on Facebook. Wang, X. ; Jia, J. ; Cai, L. ; and Tang, J. Modeling emotion influence from images in social networks. IEEE TAFFECT COMPUT’ 15, 2015. Treat each individual independently 7

Three Qs to Answer • Q 1: Does emotion contagion exist in image social

Three Qs to Answer • Q 1: Does emotion contagion exist in image social networks? • Q 2: Will social roles influence emotion contagion? • Q 3: How to better predict the emotional status of users in social networks by considering emotion contagion? 8

Q 1: Existence Q 1. 1: When your friends are happy, will you be

Q 1: Existence Q 1. 1: When your friends are happy, will you be happy? 9

Influence Q 1: Existence Q 1. 2: When predicting a user’s emotional status, will

Influence Q 1: Existence Q 1. 2: When predicting a user’s emotional status, will her friends help? Historical post logs + Previous emotion + Image features + Friends’ emotions 10 Predict User v’s emotional status at time t happiness, surprise, anger, disgust, fear, sadness

Influence Q 1: Existence Q 1. 2: When predicting a user’s emotional status, will

Influence Q 1: Existence Q 1. 2: When predicting a user’s emotional status, will her friends help? n Historical post logs o i g a t n co n o i + t o Em e g a Previous emotion m i n i r u c c o s + doe s k r o w Image features t e ial n soc + Friends’ emotions 11

Q 2: Social Role • Opinion leaders: 20% of users with largest Page. Rank

Q 2: Social Role • Opinion leaders: 20% of users with largest Page. Rank scores; • Structural hole spanners: 20% of users with lowest network constraint scores; • Others are remaining as ordinary users. on ordinary OL and SH are more influential than i t o m e n i s d l o h till users in. Sinformation diffusion [Yang’ 15]. ? n o i g a t n o c [1] Y. Yang, J. Tang, C. W. -k. Leung, Y. Sun, Q. Chen, J. Li, and Q. Yang. Rain: Social role-aware 12 information diffusion. In AAAI’ 15, 2015.

Q 2: Social Role Happy Fear X: number of friends with different social roles.

Q 2: Social Role Happy Fear X: number of friends with different social roles. Y: probability being a certain emotion. Happy Fear 13

Q 2: Social Role Happy Fear X: number of friends with different social roles.

Q 2: Social Role Happy Fear X: number of friends with different social roles. Y: probability being a certain emotion. Happy positive emotion delights friends Fear 14

Q 2: Social Role Happy Fear X: number of friends with different social roles.

Q 2: Social Role Happy Fear X: number of friends with different social roles. Y: probability being a certain emotion. Happy Fear 15

Q 2: Social Role Happy Fear X: number of friends with different social roles.

Q 2: Social Role Happy Fear X: number of friends with different social roles. Y: probability being a certain emotion. Happy Fear 16 “Emotional comfort” phenomena

Q 2: Social Role Happy Fear X: number of friends with different social roles.

Q 2: Social Role Happy Fear X: number of friends with different social roles. Y: probability being a certain emotion. Happy Fear 17 Opinion leaders are more influential on positive emotions Ordinary users are more influential on negative emotions

Q 2: Social Role Happy Fear X: number of friends with different social roles.

Q 2: Social Role Happy Fear X: number of friends with different social roles. Y: probability being a certain emotion. Happy Fear 18 Influence of opinion leaders and structural holes change faster than ordinary users.

Q 3: Model P(Y|G): Conditional probability of users’ emotional status given input data 19

Q 3: Model P(Y|G): Conditional probability of users’ emotional status given input data 19

Q 3: Model P(Y|G)=πg(. ) … g(xvt, yvt): Correlation between v’s emotion and the

Q 3: Model P(Y|G)=πg(. ) … g(xvt, yvt): Correlation between v’s emotion and the image she posts at t. 20

Q 3: Model P(Y|G)=π{g(. )h(. )} … h(yut-t’, yvt): Correlation between v’s emotion at

Q 3: Model P(Y|G)=π{g(. )h(. )} … h(yut-t’, yvt): Correlation between v’s emotion at time t and t-t’. 21

Q 3: Model P(Y|G)=π{g(. )h(. )l(. )} l(yut-1, yvt): How v’s emotion at t

Q 3: Model P(Y|G)=π{g(. )h(. )l(. )} l(yut-1, yvt): How v’s emotion at t is influenced by her friend u’s emotion at t-1. Social role sensitive parameter 22

Experimental Results Flickr dataset: 2, 060, 353 images, 1, 255, 478 users ground truth

Experimental Results Flickr dataset: 2, 060, 353 images, 1, 255, 478 users ground truth obtained by user tags Distribution of users’ emotional statuses on Flickr: happiness: 46. 2% surprise: 9. 7% anger: 8. 0% disgust: 5. 3% fear: 17. 3% sadness: 13. 5% 23

Experimental Results Baselines Methods do not consider emotion contagion: SVM, Logistic Regression (LR), Naïve

Experimental Results Baselines Methods do not consider emotion contagion: SVM, Logistic Regression (LR), Naïve Bayes (NB), Bayesian Network (BN), Gaussian Radial Basis Function Neural Network (RBF). Methods ignore social role information: CRF Our model: Role-aware 24

Experimental Results Evaluation Metrics: Precision Recall F 1 Measure 25

Experimental Results Evaluation Metrics: Precision Recall F 1 Measure 25

Experimental Results 26

Experimental Results 26

(a) Ground truth (c) Opinion leaders 27 (b) Random users (d) Structural hole spanners

(a) Ground truth (c) Opinion leaders 27 (b) Random users (d) Structural hole spanners

Conclusion • We study the interplay between users’ social roles and emotion contagions by

Conclusion • We study the interplay between users’ social roles and emotion contagions by answering 3 questions. – Does emotion contagion exist? – How social roles influence emotion contagion? – How to better predict users’ emotional status? • We propose the social role-aware contagion model and validate it on a real social network. 28

THANK YOU! Social Role-Aware Emotion Contagion in Image Social Networks Yang, Jia, Boya Wu,

THANK YOU! Social Role-Aware Emotion Contagion in Image Social Networks Yang, Jia, Boya Wu, and Jie Tang Department of Computer Science and Technology Tsinghua University Contact: Sherlock. Bourne@gmail. com http: //yangy. org