Context
The application context of this internship concern social network analysis. The theoretical context is the preference queries applied to very large databases.
The concept of preference queries has been established in the database community and was intensively studied in the last decade. These queries have dual benefits. On the one hand, they allow to interpret accurately the information needs of a given user. On the other hand, they constitute an effective method to reduce very large datasets to a small set of highly interesting results and to overcome the empty result set.
A query is personalized by applying related user preferences stored in the user���s profile.
However, with the advent of social networks such as Facebook, Twitter, Instagram, or more locally Breizbook, the user is no longer considered as an individual entity, at least more only. In this context, the user designates an interconnected social entity and is the author of significant information flow.
Scientific purpose
After a literature review, the objective of this thesis will be the development of a collaborative system for personalizing analyzes (i.e. preference queries) based on profiles of a user group.
As a first step, this would require the determination of the closest users of the current user in a given community (ring of friends on Breizbook, professional network on LinkedIn,...). Therefore the implementation of a similarity measure for comparing user profiles is required. The idea is to define the notion of similarity skyline of profile query defined by the set of profiles of the target database that are the most similar to the query in the sense of the similarity dominance relation defined in [3].
The idea is to achieve a d-dimensional comparison between user profiles in terms of d local distance (or similarity) measures and to retrieve those profiles that are maximally similar in the sense of the Pareto dominance relation.
A step of user matching will then be added upstream of preference queries [7]. The preference integration and recommandation [5] remain unchanged, but they will take as input preferences from different abd multiple users.
Then we have to assess the quality of returned profiles to the user. Indeed, the relevance of the delivered information and its adaptation to the users��� preferences are key factors for acceptance or rejection of recommendation systems. It is therefore to determine whether calculated results are consistent with expected ones. In this case, appropriate quality criteria should be defined. They can be inspired by the work in data quality [1, 4] but may also be defined by the system users themselves. The last step is therefore to qualify the matching process based on the quality of the results. This qualification could be based on the recall and precision measures. If the user is not satisfied with the results, he/she could start a new matching process using other matching criteria or by changing the settings.
Required skills
The successful candidate will hold a Master���s Degree in Computer Science (or equivalent) and demonstrate outstanding skills in the following areas (or similar):
Preference modeling
Machine Learning
Graph theory
Linked Data
Strong programming skills and English proficiency are preferential criteria. Interested applicants must prepare:
Detailed Resume
List of marks
Letter(s) of recommendation
Application
Applications should be sent (in pdf format) to Tassadit BOUADI (tbouadi@irisa.fr) and Arnaud MARTIN (arnaud.martin@irisa.fr) before Friday, June 24, 2016.
Bibliographie
[1]. D. Grigori, V. Peralta, and M. Bouzeghoub. Service retrieval based on behavioral sp��cifications and quality requirements. In WilM.P. Aalst, Boualem Benatallah, Fabio Casati, and Francisco Curbera, editors, Business Process Management, volume 3649, pages 392���397. Springer Berlin Heidelberg, 2005.
[2]. T. Hogg, Inferring preference correlations from social networks, Electronic Commerce Research and Applications, Vol. 9, pp. 29���37, 2010.
[3]. K. Abbaci, A. Hadjali, L. Lietard, and D. Rocacher. A similarity skyline approach for Handling graph queries - a preliminary report. In Proceedings of the 2011 IEEE 27th International Conference on Data Engineering Workshops, ICDEW ���11, 2011.
[4]. L. Berti-Equille, I. Comyn-Wattiau, M. Cosquer, Z. Kedad, S. Nugier, V. Peralta, S. Si-Saied Cherfi, and V. Thion-Goasdou��. Assessment and analysis of information quality : a multidimensional model and case studies. International Journal of Information Quality, 2 :300���323, 2011.
[5]. T. Bouadi, M-.O. Cordier, and R. Quiniou. Computing skyline incrementally in response to online preference modification. T. Large-Scale Data- and Knowledge-Centered Systems, 10 :34���59, 2013.
[6]. S. Dhamal and Y. Narahari. Scalable preference aggregation in social networks. In First AAAI Conference on Human Computation and Crowdsourcing, pages 42���50. AAAI, 2013.
[7]. F. Elarbi, T. Bouadi, A. Martin and B. Ben Yaghlane. Fusion de pr��f��rences pour la d��tection de communaut��s dans les r��seaux sociaux. LFA 2015
-- ======================================== Pr. Arnaud MARTIN Universit�� de Rennes 1 / IUT de Lannion UMR 6074 IRISA Rue Edouard Branly BP 30219 22302 Lannion cedex, France Phone: +(33)2.96.46.94.60 ========================================