EventMedia Live, Winner of ISWC Semantic Web Challenge, Starts New Project With Nokia Maps, Extends Architecture Flexibility
The winner of the Semantic Web Challenge at November’s International Semantic Web Conference (ISWC) was EventMedia Live, a web-based environment that exploits real-time connections to event and media sources to deliver rich content describing events that are associated with media, and interlinked with the Linked Data cloud.
This week, it will begin a one-year effort under a European Commission-funded project to align its work with the Nokia Maps database of places, so that mobile users of the app can quickly get pictures of these venues that were taken by users with EventMedia’s help.
A project of EURECOM, a consortium combining seven European universities and nine international industrial partners, EventMedia Live has its origins in the “mismatch between those sites specializing in announcing upcoming events and those other sites where users share photos, videos and document those events,” explains Raphaël Troncy, assistant professor at the EURECOM: School of Engineering & Research Center, Multimedia Communications, and one of the project’s leaders.
With the help of semantic web technologies, EventMedia Live aims to close that gap, so that people who’ve participated in events can more seamlessly relive their own past experiences, and those who weren’t there can easily live them second-hand. Another key goal is to make it easier and more serendipitous for users to discover events that are about to take place that align with their own interests. “For that,” says Troncy, “you need to maintain a lot of data.”
EventMedia Live is a hub in the Linked Data cloud obtained from three public event directories (Last.fm, Eventful, and Upcoming) and from one media directory (Flickr). “We observed that none of the event directories at the moment is exhaustive in coverage. We are not, either, but by combining multiple directories we improve coverage and give better satisfaction to the user,” Troncy says. EventMedia uses the Linked Open Description of Events (LODE) ontology to publish descriptions of events from event and media sources as Linked Data, and to map between other event-related vocabularies and ontologies.
“At the end we get a lot of information, a lot of data, representing to a single unified model, LODE, about those events,” he says. To solve the overlaps or inconsistencies that still may exist in the raw data and metadata it scrapes from the event and media directories, EventMedia Live also has developed a number of new algorithms: For instance, to know whether the same venue is being described in different instances by slightly different names, it does a mix of comparing a structured address with geo-location information.
There also are issues around event directories taking inconsistent paths to providing event or exhibition timings. Some might give a start date, for example, while others give a start and end date, which prompted EventMedia Live to develop temporal inclusion metrics to help verify whether or not sources are citing the same event.
“At the end of the day, with all those metrics we have a certain confidence that we are successfully aligned,” Troncy says. “But we tried to really minimize incorrect information…so we agreed to have a lower recall, to not find all possible alignments but make sure the ones we do publish are the good ones.” Precision, he says, is in the 98 percent range.
Media and event descriptions are enriched with background knowledge from external datasets such as DBpedia, MusicBrainz, BBC and Foursquare. Data collected from its source sites is converted into RDF triples – more than 30 million so far – providing descriptions of events using the LODE ontology and a large SKOS taxonomy of event categories.
“All the data that we extract is exposed according to 5-star Linked Data principles,” he says. “So we publish a full RDF data set that is part of the LOD cloud that links in and out to other data sets. We provide dereferencable URIs, a public SPARQL endpoint, and we also publish an API that reads more data and which the user interface is based on.”
With events and media data together, sharing the same universal data model, the end user can more quickly find out the information they want to know about, say, an upcoming concert, or one that recently took place. A visit to EventMedia Live, which contains descriptions of events and descriptions of photos illustrating events from various sources, gives the user a single entry point into aggregated, disambiguated and real-time information about when and where something takes or took place, and what the artist or experience looks or looked like, too.
In the just-launched project around Nokia Maps, which has a lot of rich information about the venues in its database (opening hours, what public transportation is available to reach the destination, and so on), EventMedia Live will step in to match photos to those places. “We want to see what we could prototype, and how we could take advantage of both rich datasets for providing more innovative apps on top of this,” says Troncy.
He points to other services that are leveraging the data EventMedia exposes, such as one in the UK that offers as open data the maps of all metro subway stations and information about access capabilities for disabled riders at each one. “They combine that with venues of events from EventMedia to provide a service to the user to say, if you want to go to this venue, we recommend that you go this way if you are disabled,” for a more seamless travel experience, he notes.
The service also is prototyping including Facebook media and event data in the mix, and continues to work on adding flexibility to its architecture, to make it easier to accommodate new event sources and to drive an expansive vision in other respects. “We are convinced there will always be a very broad range in sources publishing information, from big event directories to hyperlocal event directories,” he says.
At the moment those staging these events have to fill out the same information dozens of times to make sure their event is presented in as many catalogs as possible. Says Troncy, “We could play a role in brokering this information,” and ease that burden.