Forecasting Mood Using Smartphone and SNS Data
暂无分享,去创建一个
This poster demonstrates a mood forecasting system that forecasts tomorrow's mood based on today's data collected from SNS and smartphone sensors. Forecasting user's mood plays an important role in diverse topics such as recommendation systems. We define the mood as the variation in the feeling of a person. It can be either positive, negative or neutral. Previous works can be divided into two groups: SNS-based and sensor-based methods. For the former group, the prediction coverage depends on user's activeness on SNS. Besides, they focused on text-only posts and ignore other types of posts such us link-only, image-out posts which may contain important insights [2], [3]. For the latter group, conventional methods have been focused on collecting data using extra devices to measure body physiological signal [4] and/or on smartphone sensors [1] along with user inputs. These methods suffer from the noise introduced by the user. The research question for this study was whether it is possible to built an autonomous mood prediction system by sensing the user's smartphone sensors and SNS activities over cyber, social and physical spaces.
[1] Shivakant Mishra,et al. EmotionSensing: Predicting Mobile User Emotion , 2017, ASONAM.
[2] Diana Inkpen,et al. Monitoring Tweets for Depression to Detect At-risk Users , 2017, CLPsych@ACL.
[3] Shivakant Mishra,et al. Having Fun?: Personalized Activity-Based Mood Prediction in Social Media , 2017, Prediction and Inference from Social Networks and Social Media.
[4] Samit Bhattacharya,et al. Towards affective touch interaction: predicting mobile user emotion from finger strokes , 2015, Journal of Interaction Science.