| Institution: | Slovak University of Technology |
| Technologies used: | Ruby |
| Inputs: | programmers' recommendation prediction, social aspects, social interacion |
| Outputs: | actual programmers' satisfaction (experienced items) |
Addressed problem
As everyone, software development team and its members are strongly influenced by their social aspects – personality types, surrounding including other people etc. It is clear that these aspects strongly influence the single user mood and performance respectively. The single user preferences stored in individual user models or predicted by the recommendation algorithm do not exactly match to the final ratings after experiencing in the group or itself by the programmer. Even when we consider the programmer A and we know his/her rating for some content C1 is 5, the rating after experience and discussion within the group will probably differ. Moreover, we need to model user satisfaction after various sequences and context of recommended items, with consideration friendship or other relations’ types (which influences the final satisfaction).
Description
To overcome problems of standard satisfaction modelling (group suggestibility, other non-human aspects) we suggest a new satisfaction function, which is not based on the user's mood, but it is taking into account the group impact on specific user and other aspects respectively (user context). Because the emotional contagion is the bi-directional process, we use the spreading activation. Vertexes represent users, which have assigned initial satisfaction values, which are derived in a standard manner (collaborative/content based recommendation, prediction). The edges are present between users whose are in relationship or between related contexts, in other words between users with emotional contagion or context. Because the satisfaction does not depend only on the emotional contagion but on the time (emotion decrease over time) and previously experienced items either. Replacing the spreading activation by the real user's ratings allows us to model the satisfaction in the single-user environments, where the user is influenced only by the experienced content. Including such an approach into the recommendation process can improve the satisfaction of users – programmers in connection to obtained recommendations (based on the improvement of rating prediction and recommendations respectively).
References
Kompan, Michal: Personalized Recommendation Considering Social Aspects. In: WIKT 2011 Proceedings 6th Workshop on Intelligent and Knowledge oriented Technologies, November 24 - 25, 2011 Herľany, Slovakia. - Košice : Technická univerzita, 2011. - ISBN 978-80-89284-99-3. - S. 155-160 (in slovak)
Kompan, Michal – Bieliková, Mária: Context-based Satisfaction Modelling for Personalized Recommendations. In: The Semantic and Social Media Adaptation and Personalization workshop. Bayonne, France, IEEE. 2013 s.5 (to appear)