E-Letter‎ > ‎

STCSN E-Letter Vol.1 No.3

Sensing User Generated Content - Taking a Closer Look at the SocialSensor Project

Welcome to the STCSN E-Letter Vol.1 No.3! After STCSN published its second E-Letter earlier this year, we proudly announce its third edition, which contains fresh results from the SocialSensor project on Sensing User Generated Input for Improved Media Discovery and Experience. We are sure you will enjoy this edition, since it contains a lot of interesting thoughts and results from this European research project. As always, feel free to get in touch with us if you want to contribute to future E-Letter editions, or STCSN in general!

TABLE OF CONTENT:
  • Editorial
    (by Symeon Papadopoulos and Rene Kaiser)
  • The SocialSensor Project - Sensing User Generated Input for Improved Media Discovery and Experience
    (by Sotiris Diplaris, Symeon Papadopoulos and Yiannis Kompatsiaris)
    Abstract: Social networks have become an integral part of modern life driving more and faster communication than ever before. For politics, business and pleasure these new connections are shaping our world, and a key challenge for SocialSensor project is to enable trends, sentiments and discussions to be surfaced quickly in ways that are both relevant and useful. SocialSensor develops a new framework for enabling real-time multimedia indexing and search in the Social Web. The project moves beyond conventional text-based indexing and retrieval models by mining and aggregating user inputs and content over multiple social networking sites.  Information about the structure and activity of the users’ social network is incorporated directly into the multimedia analysis and search process via a Social Indexing procedure. Furthermore, SocialSensor enhances the multimedia consumption experience by developing novel user-centric media visualization and browsing paradigms. For example, it analyses the dynamic and massive user contributions in order to extract unbiased trending topics and events and uses social connections for improved recommendations. To achieve its objectives, SocialSensor introduces the concept of Dynamic Social COntainers (DySCOs), a new layer of online multimedia content organisation with particular emphasis on the real-time, social and contextual nature of content and information consumption. Through a DySCOs-centered media search, SocialSensor integrates social content mining, search and intelligent presentation in a personalized, context and network-aware way, based on aggregation and indexing of both UGC and multimedia Web content. This article discusses the challenges SocialSensor aspires to tackle and provides a high-level view of the system architecture, current research approaches and use case implementations.
  • Finding Newsworthy Topics on Twitter
    (by Carlos Martin, David Corney, Ayse Goker)
    Abstract: Increasingly, journalists and news organizations are using social media to discover news stories and supporting information through interactive information retrieval methods. The rate and diversity of posted messages present major challenges however, with new topics emerging continually, making it hard for the user to find what is newsworthy. We describe a novel technique to detect clusters of bursty phrases that appear in the same messages so as to identify emerging topics in a useful fashion. We demonstrate our methods by analysing tweets corresponding to events drawn from the worlds of politics (US primaries and presidential elections) and sport (the FA Cup Final). We created a user-centred “ground truth” to evaluate our methods, based on mainstream media accounts of the events. We show that our method successfully detects a range of different topics for each event and can retrieve headlines that represent each topic for the user.
  • Shortest Urban Paths or Shortcuts to Happiness?
    (by Daniele Quercia, Rossano Schifanella, Luca Maria Aiello)
    Abstract: When providing directions to a place, web and mobile mapping services are all able to suggest the shortest route. At times, however, when visiting a friend, we do not necessarily take the fastest route but might enjoy alternatives that, for example, offer beautiful urban sceneries. The goal of this work is to propose ways of automatically generating routes that are not only short but also emotionally pleasant. To quantify the extent to which urban locations are pleasant, we build a crowd-sourcing platform that shows two street scenes in London (out of hundreds), and a user votes on which one looks more beautiful, quiet, and happy. Capitalizing upon few simple engagement strategies, we have been able to collect votes from more than 3.3K individuals, and we then translate those pairwise comparisons into quantitative measures of urban perception along the three dimensions. We are also able to compute a proxy for the beauty dimension from Flickr metadata associated with more than 3.7M pictures in London. Then, to quantify the extent to which locations are popular, we crawl all Foursquare venues in the entire city. We finally arrange locations into a graph upon which we learn popular and pleasant routes. We quantitatively validate the extent to which the recommended routes are not only short (they add just a few extra minutes to the shortest routes) but also popular and emotionally-pleasing: compared to the shortest routes, they are always perceived as more pleasant, with an increase of up to 30%. We then qualitatively evaluate the recommendations by conducting a user study involving as many as 20 participants who have not only rated the recommendations but also carefully motivated their choices.
  • Social Multimedia Crawling and Search
    (by Symeon Papadopoulos, Emmanouil Schinas, Theodoros Mironidis, Yiannis Tsampoulatidis, Yiannis Kompatsiaris)
    Abstract: This paper describes an interactive system that facilitates the targeted collection, indexing and browsing of multimedia content that is shared through social media platforms. Such a system is valuable in several use cases involving the tracking of social content related to an entity of interest, e.g. topic, person, location or event. When the entity of interest is popular or trending, massive amounts of multimedia content are shared through social media, resulting in information overload for the end users, stemming mainly from the nature of social content (e.g. large amounts of near-duplicate content, spam). To this end, the proposed system supports interactive targeted crawling and indexing of social multimedia content in tandem with real-time analysis and summarization with the goal of extending the search capabilities of content seekers. The system is based on a real-time distributed architecture, including very efficient image indexing and clustering implementations. A case study is presented assessing the system on the media content shared around the #OccupyGezi events.
  • Semantic-Driven K-Walker Search in Unstructured Peer-to-Peer Networks
    • (by Xiaoqi Cao and Matthias Klusch)
    Abstract: In S2P2P, each peer maintains its observation on the semantics of received queries (demands) and data information (supplies), as well as a local view on the network topology. Each peer, when forwarding a query, disseminates its known data information to a selected set of remote peers by taking advantage of query piggybacked data. That is, each peer locally observes the query and item semantics, aka local view on the semantic overlay of the P2P network, during the k-walker search of each query it receives or issues. For routing a query, each peer, instead of merely introducing an immediate neighbor, suggests a query routing path consisting of a sequence of peers with expertise on topics which are semantically equivalent or sufficiently similar to the query topic. This is achieved by a path suggestion heuristics that iteratively applies Dijkstras algorithm in a greedy manner. Each iteration manages to detect one more expert peer and augments the current path suggestion with the shortest path from its tail to the detected expert peer. That is, inspired by the literature on gossiping in P2P networks, the optimal expertise-based routing path per query is collaboratively determined by peers as the shortest path with the maximal number of peers which are actually known to have semantic expertise for the considered query topics. In particular, the collaborative path adjustments by peers on this path are done in order to maximize the gain of the total answer for the query (within the TTL of the respective query walker). Our comparative experimental evaluation revealed that S2P2P may outperform representative approaches to semantic flooding while being at least as robust against changes in the network.
How to cite this E-Letter edition?
Symeon Papadopoulos, Rene Kaiser (ed.), "Sensing User Generated Content - Taking a Closer Look at the SocialSensor Project", IEEE Computer Society Special Technical Community on Social Networking E-Letter, vol. 1, no. 3, October 2013.