-- 发布时间：1/13/2009 8:16:00 PM
-- Report on billion triple track, semantic web challenge, ISWC 2008
Hi all,
I would like to introduce something about Semantic Web challenge, especially the new billion triple track. Although it is a bit late, I still wish you can enjoy it.
Semantic Web challenge is a contest organized every year around ISWC. The jury is composed of experts from universities and industry. In this year, it includes two tracks: open track and billion triple track. The open track aims at designing an end-user application. It encourages to use several heterogeneous data sources. It emphasizes the importance of dealing with data meaning. The billion triple track is a fresh new one. This is the first year for it. The organizer provides a big set of data sets from both semantic web search engines and high-quality ones such as DBpedia, Freebase and US Census. It requires the participants to use a significant part of the data while it does not require specially an end-user application.
The commonness between both tracks: 1) integrate, combine and dedue information, 2) give researchers an opportunity to show their work and compare it with others, 3) stimulate current research. There are several differences. In particular, the open track requires to show what the Semantic Web can provide to end users. The billion triple track deals with the scalability problem and wants to find what can be done from accumulated data sets.
There are 22 participants involved in this year's semantic web challenge. In the billion triple track, there are 9 teams in all in which eight teams are from Europe and one joint team is from our team (Apex Data & Knowledge Management Lab, SJTU) and AIFB. In the open track, 13 teams have participated. Among them, one team is from South East University, two teams are from US, one team is from BIKE lab, Korea while the others are from Europe. In particular, several companies are involved to advertise their semantic web products.
According to the systems presented in the final round of billion triple track, there are some interesting findings. 1) cloud computing is a preferred answer to tackle the scalability issue. Especially, the 1st place SemaPlorer is built based on the Amazon EC cloud while the 3nd place Marvin uses the DAS-3 distributed supercomputer. 2) You can feel the importance of reasoning not in the open track but in the billion triple track. Marvin and SOAR belong to this kind. In my opinion, large scale reasoning is not yet ready for general use but is a nice perspective. :) 3) All the systems are not started from scratch. They reuse and combin several existing "blocks". SearchWebDB is built on top of the extension of Semplore and Q2Semantic. SOAR is considered to add the reasoning feature for SWSE while the backend for i-Moco is Hexastore, a native RDF triple store. In a word, Semantic Web is now building on top of its first significant bricks.
Attached please find the slides. 
