Jennifer Zaino
SemanticWeb.com Contributor

SemanticWeb.com caught up with Yahoo Researcher Peter Mika, a former winner himself of the Semantic Web Challenge and past and current competition co-organizer (with web pioneer Jim Hendler, Tetherless World Senior Constellation Professor at Rensselaer Polytechnic Institute) to get some perspective on this year's winners, which included entries for both the Open Track and the Billion Triples Track.

SemanticWeb.com: Last year when we spoke, you mentioned that year marked the largest number of entries for the Semantic Web Challenge. How did this year compare with last in terms of numbers, and if it multiplied again, to what do you attribute the continuing growing interest by participants?

Mika: Yes, we've again seen the number growing. We have evaluated 22 submissions for the two tracks in total (the comparable number was 19 last year). The growing interest can be attributed to the introduction of a new competition (the Billion Triples Track), as well as the general growth of the conference, which was attended by over 600 participants.

The challenge also increased its share of the conference program: the applications were demonstrated during the poster and demo session, and the best applications were also presented during a plenary presentation session.

SemanticWeb.com: Explain the difference between the two categories -- the Open Track and the Billion Triples Track.

Mika: The new Billion Triples Track differs from previous years' competition and the current Open Track in that the data set the participants have been using was provided by the organization. This data set is challenging both in terms of size (one billion = 1,000 million triples) and heterogeneity. It comprises largely data collected from the public Web by existing Semantic Web search engines (Falcon, Sindice, Swoogle, SWSE, Watson); thus, dealing with this data set was expected to be much more difficult [for reasons including] inherent quality issues. As we have been uncertain about what the participants would be able to achieve with such a difficult data set, the goal of this Track was defined loosely "as adding value to the data," either by building an end-user interface or by performing computations on the data set that allows building other applications on top of the data. In fact, we have received submissions of both kinds.

SemanticWeb.com: For the Open Track awards, first prize was awarded to paggr, second to DBpedia Mobile, and third to HeatlhFinland. Was there anything that these applications shared in common -- and also, what about each application were the unique features that helped it take each honor?

Mika: The applications are actually quite different, showing the versatility of semantic technologies. (We have long observed that one of the difficulties of showcasing semantic technologies is that the technology itself is not always visible from the outside, nor is it clear if it should be.) The very fact that such complex applications could be assembled as a personal project in many cases (paggr, DBpedia Mobile), and in a relative short time, speaks volumes to the maturity of readily available tools and APIs for building semantic applications.

One theme that cuts across the two tracks was the increasing interest in mobile applications (DBpedia Mobile, I-MoCo, IYOUIT).
What made paggr stand out was the incredible end user experience. Paggr is an interactive AJAX-based application where users build personal information spaces by organizing and integrating distributed web information coming from RSS feeds, microformats, SPARQL endpoints etc.

DBpedia Mobile provided a novel and compelling functionality as a travel companion where one can see neighboring objects described by Wikipedia pages and also upload new information (location, photos and reviews of objects).

HealthFinland is the result of a large Finnish national effort to create a comprehensive portal of health information based on content that is contributed by multiple partners and described using a shared ontology.

SemanticWeb.com: For the Billion Triples Track, I see that first prize went to SemaPlorer, second to SearchWebDB, and third to MaRVIN. Again, let me ask what about each application were the unique features that qualified it to take the top prizes?

Mika: On the technical side, SemaPlorer showed distributed querying capabilities over significant parts of the dataset, and also provided a very compelling end-user experience. SearchWebDB innovated on the query evaluation side: it matched complex, multi-word keyword queries to a summary of the graph to obtain the most likely interpretation of the query. MaRVIN was the most high risk and high potential experiment of all: the authors performed RDF reasoning on the entire dataset using parallel computing methods.

SemanticWeb.com: What surprised you the most about the entries the committee saw this year-- especially as it relates to the Billion Triples Challenge? I know last year we talked about how you wanted the Billion Triples challenge to push the scale -- how did the outcome square with your hopes on that front?

Mika: We did achieve our initial goal of attracting more high-risk projects that go toward implementing semantic technologies in the open web scenario. We found the new task was not easy: despite popular claims that a billion triples is no obstacle these days, we know that many participants worked very hard to complete their projects, and there were no doubt countless projects that have been abandoned along the way.

Nevertheless, we hope that next year even more projects make it to the final, reaching new milestones in using semantic technologies in the very difficult Web scenario.

SemanticWeb.com: So, just as the Billion Triples was the big new thing this year, what might you all be thinking about to make next year's challenge even more interesting?

Mika: We are planning to run both the Open Track and the new Billion Triples Track in a similar form. We plan to keep the size of the Billion Triples dataset but change its composition slightly to reflect even more closely the characteristics of Web data. No doubt, as in the case of previous competitions, the bar will be set even higher next year by the participants themselves. I will be joined by Chris Bizer from the Freie Universitat Berlin as co-organizer of next year's event.