MediaMixer partner EURECOM not only contributes directly to future media technology, co-authoring the Media Fragments URI specification and working on media fragments and semantic multimedia implementations, but has been demonstrating the technologies’ value at leading computing conferences, namely the World Wide Web Conference (WWW) 2013 this past May in Brazil as well as the Extended Semantic Web Conference (ESWC) 2013 shortly thereafter in France.
At the Linked Media (LiMe2013) workshop at WWW2013, EURECOM presented “Enriching Media Fragments with Named Entities for Video Classification”. In this work, we propose a framework which classifies video according to textual features such as named entities spotted from subtitles, and temporal features such as the duration of media fragments where entities are spotted. We implement four automatic machine learning algorithms for multi-class classification problems named Logistic Regression (LG), K-Nearest Neighbour (KNN), Naive Bayes (NB) and Support Vector Machine (SVM). The results show that this approach is promising in the context of online videos classification.
EURECOM also organised a workshop at WWW2013 entitled “Web of Linked Entities (WoLE)“, which transparently connects the World Wide Web (WWW) and the Giant Global Graph (GGG) using methods from Information Retrieval (IR), Natural Language Processing (NLP) and Database Systems (DB). The workshop has attracted 23 submissions from all over the world on which the program committee has selected 9 papers. The workshop also features two keynotes from Peter Mika (Yahoo!) for a talk entitled Entity Search on the Web and from Icaro Medeiros (Globo.com) for a talk entitled Linked Data at Globo. Finally, David Graus won the workshop challenge with the SemanticTED demo.
At another workshop, on real-time analysis and mining of social streams (RAMSS), EURECOM has given an invited talk: MediaFinder: Collect, Enrich and Visualize Media Memes Shared by the Crowd which introduces the MediaFinder tool developed for finding social media content shared on social networks and organize automatically this content for creating stories. See also the slides at http://www.slideshare.net/troncy/mediafinder-collect-enrich-and-visualize-media-memes-shared-by-the-crowd
EURECOM participated in the MSM (Making Sense of Microposts) challenge and sponsored by eBay and finished 2nd out of the 22 candidates with its NERD tool. See also the presentation athttp://www.slideshare.net/giusepperizzo/learning-with-the-web-spotting-named-entities-on-the-intersection-of-nerd-and-machine-learning
At ESWC2013, EURECOM presented the demo paper entitled Tracking and Analyzing The 2013 Italian Election which builds on top of the MediaFinder tool and enables to repeat the same query over a longer period of time in order to monitor a particular event and its dynamics. You can relive the demo at http://mediafinder.eurecom.fr/story/elezioni2013 where the Italian election has been used as a use case for visualizing how social media content have been shared during the 7 days following the election day.