Intervention-Based Clustering

Title: Intervention-Based Clustering
Dates: 2016 – 2020
Researchers: Igor Kulev and Pearl Pu
Keywords: activities of daily living, machine learning, timeseries analysis, clustering

While an increase in physical activity levels has been shown to be the most effective and important intervention strategy to improve health in elderly populations, it is notoriously difficult to increase physical activity, mobilization, and rehabilitation among older adults. We propose a lifestyle recommender system to help senior adults with sedentary habits to become more active. In that context methods are developed to cluster existing users based on how a given intervention of behavior change affects their current habits modelled as time series data.

Sponsor: REACH
Links: Main Page

Behavior Recommender Systems

Title: Behavior Recommender Systems
Dates: 2013-2017
Researchers: Onur Yuruten and Pearl Pu
Keywords: Activities of daily living, activity recognition, data mining, time series analysis, statistical analysis, behavior recommenders
Abstract: Sedentary lifestyles influence the onset of many serious health problems. Healthy behavior change is an arduous task, but behavior recommenders can help their users achieve it. This novel type of recommender system can suggest its users to pair up with an exercise partner, or propose a day-by-day plan to increase their physical activeness. We quest for understanding the factors that influence well-beings of people, and develop a behavior recommender system that exploits these factors to provide suggestions for healthier lifestyles. Toward this end, we analyse various aspects of activities of daily living, such as physical exercises, work, sleep, social interactions, and daily routines. We utilize state-of-the-art timeseries analysis methods to explore trends, repetitive patterns, regularities and correlations from data such as bluetooth proximity, cell-tower locations, calorie expenditure, and other various information collected from pervasive, wearable sensors. We also show that traditional factor analysis methods can also be applied to sensor data.
Sponsor: Nano-tera
Links: Main page


Title: Dystemo: Distant Supervision for Emotion Recognition in Tweets
Dates: 2013 – 2016
Researchers: Valentina Sintsova and Pearl Pu
Keywords: emotion recognition, social media analysis, distant supervision, twitter

Emotion recognition in text has become an important research objective. It involves building classifiers capable of detecting human emotions for a specific application, e.g. analyzing reactions to product launches, monitoring emotions at sports events, or discerning opinions in political debates. Most successful approaches rely heavily on costly manual annotation. To alleviate this burden, we propose a distant supervision method—Dystemo—for automatically producing emotion classifiers from tweets labeled using existing or easy-to-produce emotion lexicons. The goal is to obtain emotion classifiers that work more accurately for specific applications than available emotion lexicons.

Links: Main Page

Emotion Recognition of Influential Users

Title: Emotion Recognition of Influential Users
Dates: 2013 – present
Researchers: Lionel Martin and Pearl Pu
Keywords: emotion recognition, influence detection, prediction, social media analysis, natural language processing, Geneva Emotion Wheel

People increasingly rely on other consumers’ opinion to make online purchase decisions. Amazon alone provides access to millions of reviews, risking to cause information overload to an average user. Recent research has thus aimed at understanding and identifying reviews that are considered influential. Most of such works analyzed the structure and connectivity of social networks to identify influential users. We believe that insight about influence can be gained from analyzing the affective content of the text as well as affect intensity. We employ text mining to extract the emotionality of thousands of reviews in different domains to investigate how those influential users behave. We analyze whether texts with words and phrases indicative of a writer’s emotions, moods, and attitudes are more likely to trigger a genuine interest compared to more neutral texts.

Sponsor: Swiss National Science Foundation
Links: Main Page

Visual Analytics for Pervasive Sensor Data

Title: Visual Analytics for Pervasive Sensor Data
Dates: 2013
Researchers: Onur Yuruten and Pearl Pu
Keywords: Activities of daily living, visual analytics, visualisation

Advances in sensor technology have enabled the deployment of pervasive healthcare systems in increasingly larger scales. There is a ‘quantified self’ movement, which involves using wearable sensors to quantify external activities of daily living (ADL) and internal metabolism. Many studies show that these systems can improve well-being of their users via presenting such information with self-reflection tools. However, there is a lack of visualization tools to aid analytical investigations over the data collected from these systems, which could improve our understanding on factors that influence well-being, and guide the design for better wellness management systems.

We develop a visual analytics tool to discover the interpersonal and intrapersonal patterns of ADLs. With this tool, we aim to observe possible influencing factors of well-being, which could be use to construct or validate models for statistical analysis.

This project is part of a larger framework of lifestyle recommenders research, and proceeds in parallel with “mining and analysing factors of well-being project.

Sponsor: Nano-Tera
Links: Main Page

HealthyTogether: Exploring Social Incentives for Mobile Fitness Applications

Title: HealthyTogether: Exploring Social Incentives for Mobile Fitness Applications
Dates: 2012 – present
Researchers: Yu Chen and Pearl Pu
Keywords: Mobile fitness applications; gamification; competition; cooperation, group interaction.
Abstract: A crucial element in many mobile fitness applications is gamification that makes physical activities fun. While many methods focus on competition and individual users’ interaction with the game, the aspect of social interaction and how users play games together in a group remains an open subject. To investigate these issues, we developed a mobile game, HealthyTogether, to understand how users interact in different group gamification settings: competition, cooperation, or hybrid. We describe the results of a user study involving 18 dyads (N=36) over a period of two weeks. Results show that users significantly enhanced physical activities using HealthyTogether compared with when they exercised alone by up to 15%. Among the group settings, cooperation (21% increase) and hybrid (18% increase) outperformed competition (8% increase). Additionally, users sent significantly more messages in cooperation setting than hybrid and competition. Furthermore, physical activities are positively correlated with the number of messages they exchanged.
Sponsor: Swiss National Science Foundation
Links: Main Page

Recognizing Emotions in Olympic Tweets

Title: Fine-Grained Emotion Analysis in the Tweets about Olympic Games
Dates: 2012-current
Researchers: Valentina Sintsova and Pearl Pu
Keywords: Emotion Recognition, Social Media Analysis, Emotion Visualization, Human Computation
Description: Social media platforms such as Twitter or Sina Weibo have become a common way for people to share opinions and emotions. Sports events are traditionally accompanied by strong emotions and the 2012 summer Olympic Games in London were not an exception. This project aims to develop tools for analysis of the emotions expressed in social media during the Olympic Games or other popular sports events. We use the 20 fine-grained emotion categories of GEW to allow users to distinguish the personal reactions with more details. We work both on the automatic extraction of those emotions in the tweets and on their compact visualization.
Sponsor: Swiss National Science Foundation
Links: Main Page (images, visualization demo and video, AMT task demo, emotion recognition poster and paper)


Title: Empatheticons: Designing Emotion Awareness Tools for Group Music Experience
Year: 2012
Researchers: Yu Chen and Pearl Pu
Keywords: Group music player, user experience, social interaction, group influence, emotion awareness, emotion presentation
Description: Online group music players provide music playlists for a group of people. In such systems, group members sometimes listen to group songs individually. How to create the sense of connectedness for members who are listening to music in different time and locations is understudied. In this project, we investigate the roles of emotion awareness tools and how they may enable connectedness. We first describe the design of empatheticons, a set of kinetic emotion representations. We then show that they allow users to represent, annotate, and visualize group members’ emotions in GroupFun, a group music player. An in-depth user study (N = 18) with GroupFun demonstrates that empatheticons enhance users’ perceptions of the connectedness (immediacy) and familiarity (intimacy) with each other in a non-collocated and asynchronous setting. Furthermore, users’ emotion annotation for group songs can be influenced by other group members. We conclude with design implications and directions for future research.
Sponsor: Swiss National Science Foundation
Links: main page (images, video, demo, paper)


Title: ResQue: A User-Centric Evaluation Framework for Recommender  Systems
Dates: 2010 – 2011
Researchers: Rong Hu, Li Chen and Pearl Pu
Keywords: Recommender systems, quality of user experience, e-Commerce recommender, post-study questionnaire.

This research was motivated by our interest in understanding the criteria for measuring the success of a recommender system from users’ point view. Even though existing work has suggested a wide range of criteria, the consistency and validity of the combined criteria have not been tested. In this paper, we describe a unifying evaluation framework, called ResQue (Recommender systems’ Quality of user experience), which aimed at measuring the qualities of the recommended items, the system’s usability, usefulness, interface and interaction qualities, users’ satisfaction with the systems, and the influence of these qualities on users’ behavioral intentions, including their intention to purchase the products recommended to them and return to the system. We also show the results of applying psychometric methods to validate the combined criteria using data collected from a large user survey. The outcomes of the validation are able to 1) support the consistency, validity and reliability of the selected criteria; and 2) explain the quality of user experience and the key determinants motivating users to adopt the recommender technology. The final model consists of thirty two questions and fifteen constructs, defining the essential qualities of an effective and satisfying recommender system, as well as providing practitioners and scholars with a cost-effective way to evaluate the success of a recommender system and identify important areas in which to invest development resources.

Sponsor: Swiss National Science Foundation
Links: Main page


Title: Designing emotion annotation interface for group recommender systems
Dates: 2011
Researchers: Yu Chen and Pearl Pu
Keywords: Emotion awareness; emotion annotation; group recommender systems; group influence 
Abstract: Group recommender systems help users to find items of interest collaboratively. Support for such collaboration has been mainly provided by interfaces that visualize membership awareness, preference awareness and decision awareness. However, these mechanisms do not address group influence issues: how members may affect each other. In this paper, we investigate the roles of emotion awareness interfaces and how they may enable positive group influence. We first describe the design process behind an emotion annotation tool, which we call CoFeel. We then show that it allows users to annotate and visualize group members’ emotions in GroupFun, a group music recommender. An in-depth pilot user study with GroupFun suggests that emotion awareness features have the potential to enable group influence on preferences among group members. Most importantly, users reported CoFeel as a promising tool to annotate group recommended items in terms of accuracy and engagement. Finally, we provide design implications for emotion awareness interfaces in group recommender systems.
Sponsor: Swiss National Science Foundation
Links: Main Page