Welcome to the second newsletter of the European project SocialSensor. Co‐funded by the EU 7th Framework Programme, SocialSensor is an Integrated Project that started its activities on the 1st of October 2011.
We are happy to announce SocialSensor successfully passed its first annual review last February in Brussels. In this issue we present a synthesis of the principal results, news items and highlights of the project. Notably, the prototype GUI for the news application, which was developed in collaboration with journalists from BBC and Deutsche Welle. Please also checkout the Alethiometer system, which measures the credibility of information retrieved from social networks. From an academic point of view the project is flourishing as well with a best paper honorable mention at this years Multimedia Modelling conference, and successful participation to the internationally reputed MediaEval benchmark initiative.
Finally, we’d like to invite you to register as a member of the SocialSensor User Group (SUG), which is a community of interest of relevant stakeholders from the Social Web. SUG members have the opportunity to actively engage with the project, such as evaluating prototypes and project results. To become a member please send a mail to email@example.com.
Social Sensor News Interface
We have been working hard to come up with an interface that showcases the best of social media for news professionals – in real time.
After a number of iterations with the designer, we have come up with a simple clean look that shows breaking trends in networks like Facebook and Twitter – but also a topic based view that brings together content from multiple social networks.
There are also powerful features that will eventually allow journalists to set their own parameters around a story so that our tool can search for relevant content from social media sources that they wish to track. Finally, we have included some new ways of evaluating a contributor or piece of content, that we hope will enable journalists to make better decisions. We call this the ‘Alethiometer’ – after the Greek word for truth
Over the next few months we’ll be integrating these designs with the development modules in our live environment. And then we’ll be ready for testing – when we can start to get valuable feedback from working journalists.
Alethiometer: Verifying SocialSensor Content
Alethiometer, a software module attempting to measure the credibility of information coming from any social source, is currently being prepared for incorporation into the SocialSensor framework. ‘Alethia’ is the Greek word for ‘truth’ and, as discovering the truth requires multiple views, we will verify information through what we call the ‘triple-C’ structure, which consists of 3 criterion categories:
(a) Contributor – We will provide data about the source of information and try to calculate the trust level, reputation and influence of this source;
(b) Content – We will try to show whether the content is reliable by checking things like the language used, the history (has it been posted before?) and possible manipulations (has an image been altered?);
(c) Context – We will try to indicate whether the ‘what’, ‘when’ and ‘where’ are interrelated.
All criteria have been quantified, weighted and combined together to provide a single metric that will give a sense of truth to the reader. At the same time, all evidence that leads to this result will be provided, so that the user can also subjectively decide on how credible every piece of information is. At the current stage of work, most of the theoretical work regarding the identification of methods for quantifying our criteria, and for combining them into a single metric is complete, and we are now proceeding to the implementation phase.
Screenshot of the future AlethioMeter Homepage
Topic Detection in Social
Online content and interactions spontaneously generated by the social media crowd contain rich and timely information about real-world events. However, the amount of data available and the vastness of the user base pose a big challenge on information filtering and organization to drill down to relevant topics and events. The task of trending topic detection is the fundamental building block to monitor and summarize the information originating from social sources, but the variety of possible methods and related variables may considerably affect the quality of results. To shed light on this aspect, several partners of SocialSensor, namely CERTH, ALBLF, Yahoo and City, explored how the textual content of posts, the volume of terms in time, and the nature of the event of interest affect the performance of topic detection. To do so, they implemented and compared six topic detection methods and proposed a temporal metric to refine the quality of detected topics. A comparative study was conducted on two Twitter datasets, one on Super Tuesday
and one on FA Cup
, differing with regard to their time scale and topic churn rate.
The collected datasets along with ground truth, created by the partners and reference implementations of the proposed methods, will be made publicly available.
Figures: Tweets in time for Super Tuesday and FA Cup datasets and associated ground truth topics
Best Paper Honorable Mention in MMM 2013
The paper "Semi-supervised concept detection by learning the structure of similarity graphs" by Symeon Papadopoulos, Christos Sagonas, Yiannis Kompatsiaris and Athena Vakali won a Best Paper Honorable Mention in the International Conference in Multimedia Modeling that took place in Huangshan, China between 7 and 9 January 2013. The paper was selected out of 96 papers, 43 oral presentations and 53 poster papers that were accepted by the conference (27% acceptance rate for oral, 44% overall acceptance rate). The paper presented a new technique for the automatic detection of semantic concepts, such as “sky”, “weather” and “fireworks”, in images. This is a well-known problem in multimedia research (frequently termed as multimedia annotation or concept detection), which is valuable in the context of SocialSensor since it enables multimedia filtering based on the detected concepts, i.e. it enables more efficient browsing and search in large multimedia collections.
You can find an online presentation of the paper on slideshare.
ThessFest Applications for 53rd Thessaloniki International Film Festival
The SocialSensor Infotainment use case has supported the 53rd Thessaloniki International Film Festival (TIFF) with its technologies. CERTH has released an update of the ThessFest, the official festival mobile application. The applications are available for free on iOS and Android. ThessFest features the complete film schedule, as well as real-time updates from Twitter on what happens in the Festival. CERTH has also build the ThessFest website an online source providing statistics on the usage of the mobile application with respect to TIFF. It features statistics such as most popular and top rated films and most active Twitterers.
ThessFest Presentation in DSLL Living Labs Event
ThessFest, the SocialSensor application built for the Thessaloniki International Film Festival, was presented at the event "Mobile Apps for Use in the City", a Digital Spaces Living Lab event about smart city mobile apps, on January 27 2013, at Vivacom ArtHall, Sofia. You can find the presentation slides online on slideshare.
Co-organisation and Participation to SED task at MediaEval 2012
SocialSensor successfully participated in the Social Event Detection task of the MediaEval 2012 workshop. MediaEval, a workshop focused on focused evaluation tasks with respect to multimedia analysis, took place in Pisa, Italy (4-5 October 2012). CERTH took part in the Social Event Detection (SED) task that involved three challenges: (a) detection of technical events in Germany, (b) soccer events in two major European cities, (c) detection of Indignados events in Madrid. All challenges were defined on a large set of over 160,000 images from Flickr. A more detailed definition of the task is available.
In all the challenges, CERTH achieved above median performance using a versatile and flexible approach developed within the SocialSensor project. The approach employed by CERTH is described in the following article, while the presentation given at the workshop is available on slideshare. We have also made available the implementation of the approach in the form of a Java library.
The 2013 Social Event Detection Task at the MediaEval Benchmark
For a second consecutive year SocialSensor is supporting the 2013 Social Event Detection Task, part of the MediaEval Benchmark Initiative for Multimedia Evaluation. This task requires participants to discover social events and organize the related media items in event-specific clusters, within a collection of Web multimedia. The task is of interest to researchers in the areas of information retrieval, multimedia content analysis, social media analysis, event detection and event-based multimedia indexing.
For a second consecutive year SocialSensor will support the Workshop on Mining Social Network Dynamics (MSND2013), which this year will be hosted by the WWW 2013 Conference in Rio de Janeiro, Brazil, May 13, 2013.
Workshop on Mining Social Network Dynamics (MSND 2013)
The research leading to these results has received funding from the European Community's Seventh Framework Programme (FP7/2007-2013) under grant agreement 287975