Shlomo Berkovsky

Intelligent food planning: personalized recipe recommendation

As the obesity epidemic takes hold across the world many medical professionals are referring users to online systems aimed at educating and persuading users to alter their lifestyle. The challenge for many of these systems is to increase initial adoption and sustain participation for sufficient time to have real impact on the life of its users. In this work we present some preliminary investigation into the design of a recipe recommender, aimed at educating and sustaining user participation, which makes tailored recommendations of healthy recipes. We concentrate on the two initial dimensions of food recommendations: data capture and food-recipe relationships and present a study into the suitability of varying recommender algorithms for the recommendation of recipes.

Physical activity motivating games: virtual rewards for real activity

Contemporary lifestyle has become increasingly sedentary: little physical (sports, exercises) and much sedentary (TV, computers) activity. The nature of sedentary activity is self-reinforcing, such that increasing physical and decreasing sedentary activity is difficult. We present a novel approach aimed at combating this problem in the context of computer games. Rather than explicitly changing the amount of physical and sedentary activity a person sets out to perform, we propose a new game design that leverages user engagement to generate out of game motivation to perform physical activity while playing. In our design, players gain virtual game rewards in return for real physical activity performed. Here we present and evaluate an application of our design to the game Neverball. We adapted Neverball by reducing the time allocated to accomplish the game tasks and motivated players to perform physical activity by offering time based rewards. An empirical evaluation involving 180 participants shows that the participants performed more physical activity, decreased the amount of sedentary playing time, and did not report a decrease in perceived enjoyment of playing the activity motivating version of Neverball.

Features Predicting Weight Loss in Overweight or Obese Participants in a Web-Based Intervention: Randomized Trial

Background Obesity remains a serious issue in many countries. Web-based programs offer good potential for delivery of weight loss programs. Yet, many Internet-delivered weight loss studies include support from medical or nutritional experts, and relatively little is known about purely web-based weight loss programs. Objective To determine whether supportive features and personalization in a 12-week web-based lifestyle intervention with no in-person professional contact affect retention and weight loss. Methods We assessed the effect of different features of a web-based weight loss intervention using a 12-week repeated-measures randomized parallel design. We developed 7 sites representing 3 functional groups. A national mass media promotion was used to attract overweight/obese Australian adults (based on body mass index [BMI] calculated from self-reported heights and weights). Eligible respondents (n = 8112) were randomly allocated to one of 3 functional groups: information-based (n = 183), supportive (n = 3994), or personalized-supportive (n = 3935). Both supportive sites included tools, such as a weight tracker, meal planner, and social networking platform. The personalized-supportive site included a meal planner that offered recommendations that were personalized using an algorithm based on a user’s preferences for certain foods. Dietary and activity information were constant across sites, based on an existing and tested 12-week weight loss program (the Total Wellbeing Diet). Before and/or after the intervention, participants completed demographic (including self-reported weight), behavioral, and evaluation questionnaires online. Usage of the website and features was objectively recorded. All screening and data collection procedures were performed online with no face-to-face contact. Results Across all 3 groups, attrition was high at around 40% in the first week and 20% of the remaining participants each week. Retention was higher for the supportive sites compared to the information-based site only at week 12 (P = .01). The average number of days that each site was used varied significantly (P = .02) and was higher for the supportive site at 5.96 (SD 11.36) and personalized-supportive site at 5.50 (SD 10.35), relative to the information-based site at 3.43 (SD 4.28). In total, 435 participants provided a valid final weight at the 12-week follow-up. Intention-to-treat analyses (using multiple imputations) revealed that there were no statistically significant differences in weight loss between sites (P = .42). On average, participants lost 2.76% (SE 0.32%) of their initial body weight, with 23.7% (SE 3.7%) losing 5% or more of their initial weight. Within supportive conditions, the level of use of the online weight tracker was predictive of weight loss (model estimate = 0.34, P < .001). Age (model estimate = 0.04, P < .001) and initial BMI (model estimate = -0.03, P < .002) were associated with frequency of use of the weight tracker. Conclusions Relative to a static control, inclusion of social networking features and personalized meal planning recommendations in a web-based weight loss program did not demonstrate additive effects for user weight loss or retention. These features did, however, increase the average number of days that a user engaged with the system. For users of the supportive websites, greater use of the weight tracker tool was associated with greater weight loss.

Putting things in context: Challenge on Context-Aware Movie Recommendation

The Challenge on Context-Aware Movie Recommendation (CAMRa) was conducted as part of a join event on Context-Awareness in Recommender Systems at the 2010 ACM Recommender Systems conference. The challenge focused on three context-aware recommendation tasks: time-based, mood-based, and social recommendation. The participants were provided with anonymized datasets from two real world online movie recommendation communities and competed against each other for obtaining the highest recommendation accuracy for each task. The datasets contained contextual features, such as mood, plot annotation, social network, and comments, normally not available in movie recommendation datasets. Over 40 teams from 20 countries participated in the challenge. Their participation was summarized by 10 papers accepted to the CAMRa workshop.

Social networking feeds: recommending items of interest

The success of social media has resulted in an information overload problem, where users are faced with hundreds of new contributions, edits and communications at every visit. A prime example of this in social networks is the news or activity feeds, where the actions (friending, commenting, photo sharing, etc) of friends on the network are presented to users in order to inform them of the network activity. In this work we endeavour to reduce the burden on individuals of identifying interesting updates in social network news feeds by automatically identifying and recommending relevant items to individuals where item relevance is based on the observed interactions of the individual with the social network. The results of our offline study show that combining short term interest models, exploiting previous viewing behavior of users, and long-term models, exploiting previous viewing of network actions, was the best predictor of feed item relevance.

Designing games to motivate physical activity

Engagement with computer games causes children and adolescent users to spend a substantial amount of time at sedentary game playing activity. We hypothesise that this engagement can be leveraged to motivate users to increase their amount of physical activity. In this paper, we present a novel approach for designing computer games, according to which the users physical activity reinforces their game character. This way the users are seamlessly motivated to perform physical activity while maintaining their enjoyment of playing the game.

Group recommendation in context

The 2011 Challenge on Context-Aware Movie Recommendation (CAMRa2011) was held in conjunction with the Fifth ACM Conference on Recommender Systems (RecSys2011). The challenge focused on group-based recommendation for households, as well as identification of household members who had rated specific movies. The participants were provided with anonymized datasets from a real world online movie recommendation community and competed against each other for obtaining the highest recommendation accuracy. Over 45 teams from 23 countries participated in the challenge. Their participation was summarized by the nine papers accepted to the CAMRa2011 workshop. This paper presents an overview of the tracks and datasets of CAMRa2011.

Recommending food: reasoning on recipes and ingredients

With the number of people considered to be obese rising across the globe, the role of IT solutions in health management has been receiving increased attention by medical professionals in recent years This paper focuses on an initial step toward understanding the applicability of recommender techniques in the food and diet domain By understanding the food preferences and assisting users to plan a healthy and appealing meal, we aim to reduce the effort required of users to change their diet As an initial feasibility study, we evaluate the performance of collaborative filtering, content-based and hybrid recommender algorithms on a dataset of 43,000 ratings from 512 users We report on the accuracy and coverage of the algorithms and show that a content-based approach with a simple mechanism that breaks down recipe ratings into ingredient ratings performs best overall.

Activmon: encouraging physical activity through ambient social awareness

In this paper we discuss the use of low-complexity interfaces to encourage users to increase their level of physical activity. We present ActivMON - a wearable device capable of representing a users individual activity level, and that of a group, using an ambient display. We discuss the results of a preliminary usability evaluation of ActivMON.

Catch-up TV recommendations: show old favourites and find new ones

Web-based catch-up TV has revolutionised watching habits as it provides users the opportunity to watch programs at their preferred time and place, using a variety of devices. With the increasing offer of TV content, there is an emergent need for personalised recommendation solutions, which help users to select programs of interest. In this work, we study the watching patterns of users of an Australian nation-wide catch-up TV service provider and develop a suite of approaches for a catch-up recommendation scenario. We evaluate these approaches using a new large-scale dataset gathered by the Web-based catch-up portal deployed by the provider. The evaluation allows us to compare the performance of several recommenders that address the discovery of both TV programs already watched by users and new programs that users may find relevant.