Extracting Knowledge from Online Social Networks and their Content
Speaker: Panayiotis Tsaparas, Associate Professor, University of Ioannina
Date: 20 May 2016 Time: 12:00 - 14:00
Location: Seminar room STEP-C (1st floor)
Host: Irini Fundulaki, ISL, ICS-FORTH


The past decade has been marked by the explosion of social networks in the online world. Such networks contain information either in their structure, or in the content posted on them by their users. Extracting this information is valuable for understanding these networks, but also for various applications. In this talk, we will consider two problems in the area of Online Social Network mining.

The first problem we consider regards the understanding of the strength of relationships in online social networks. To this end we will use the principle of Strong Triadic Closure which states that it is not possible for two individuals to have a strong relationship with a common friend and not know each other. We consider the problem of labeling the ties of a social network as strong or weak so as to enforce the Strong Triadic Closure property. We formulate the problem as a novel combinatorial optimization problem, and we study it theoretically. Although the problem is NP-hard, we are able to identify cases where there exist efficient algorithms with provable approximation guarantees. Experiments on real data indicate that our labeling agrees with practical measures of tie strength.

The second problem we consider regards the summarizing of micro-reviews posted on Location Based Social Networks such as FourSquare. Micro-reviews are bite-size reviews (usually under 200 characters), that capture the immediate reaction of users. They are rich in information, concise, and to the point. However, the abundance of micro-reviews and their telegraphic nature makes it difficult for users to extract the useful information from them, especially when going through them on a mobile device.  We consider the problem of producing a summary for a micro-review collection for an entity that is representative, compact, and readable.  We define the problem, as the problem of synthesizing a new ``review'' using snippets of full-text reviews. To balance compactness and representativeness, we formulate our problem using the Minimum Description Length principle.  We propose approximation and heuristic algorithms, and we evaluate them on real-life data collected from Foursquare and Yelp. We demonstrate that our summaries outperform individual reviews, as well as existing summarization approaches.


Panayiotis Tsaparas completed his undergraduate studies at Computer Science Department of University of Crete, Greece in 1995. He continued his graduate studies at University of Toronto, where he received his M.Sc. , and Ph.D degree, in 2003, under the supervision of Allan Borodin. His Ph.D. thesis was on the study of Link Analysis Ranking algorithms for the Web. After graduation, he worked as a post-doctoral fellow at University of Rome, “La Sapienza”, as a researcher at University of Helsinki, as a visiting researcher at NIH-NCBI, and most recently as a researcher at Microsoft Research. Since 2011 he is joined the Department of Computer Science and Engineering of University of Ioannina, where he is now an Associate Professor.  His research interests include Social Network Analysis, Algorithmic Data Mining, Web Mining and Information Retrieval. He is an active member of the community and he has served several times as a PC member for conferences such as KDD, WWW, WSDM, VLDB, ICDE, ECML/PKDD, ICDM, SDM, and as reviewer for journals such as TKDE, TWEB, CACM, TODS, SIDMA, TOIT. He has served in three NSF panels, and twice as a reviewer for the Microsoft Faculty Fellowship Award program, and as a reviewer for Hellenic Research Foundation. During his tenure at Microsoft he received 3 technology transfer awards for successful transfer of research results to product groups. He has published 44 papers in peer-reviewed conferences and journals, and has filed for 12 patents, 8 of which have been awarded. According to Google Scholar he has a total of 2800 citations, and h-index 23. His Erdos Number is 3.

© Copyright 2007 FOUNDATION FOR RESEARCH & TECHNOLOGY - HELLAS, All rights reserved.