Significant technological breakthrough in Galway
04.05.2007
DERI Galway achieved a significant technological breakthrough in Semantic Web technology: DERI's Semantic Web Search Engine is able to answer queries with more than 7 Billion RDF Statements in fractions of a second.
Semantic World Record
DERI's Semantic Web Search Engine is able to answer queries with more than 7 Billion RDF Statements (an RDF statement is the entity that makes the Semantic Web semantic) in fractions of a second - the largest number reported so far anywhere in the world. Possible application areas include Social Network Applications and Analysis, eHealth applications, Web Search, location based services, financial search and many more.
"The importance of this breakthrough can not be overestimated" says Prof. Stefan Decker, director of DERI Galway, "Now we can really make the semantic web scale - these results enable us to create web search engines that really deliver answers instead of links and is able to combine information from the Web. For example the engine can list all partnerships of a company even if there is no single web page that lists all of them."
Andreas Harth and Aidan Hogan, key researchers on the Semantic Web Search Engine project, have been working on the project for about three years are excited about the prospects: "These were the fruits of hard labor" says Andreas Harth, "I am excited about the prospects ahead. We are currently working on realizing inferencing - making the web truly intelligent - and we have results already."
Short Interview with Andreas Harth
Andreas Harth one of the key researchers on the Semantic Web Search Engine project talked with the Semantic Web School about SWSE:
SWS: What kind of data did you use for this world record?
Harth: We used the Lehigh Univeristy Benchmark LUBM(50000). Thatīs around 75 times of the amount of the DBpedia dataset. The structure and characteristics of the datasets might be slightly different (the synthetic dataset uses a different ontology than the wikipedia dataset), but from a scaling perspective that shouldn't matter a lot.
SWS: What kind of reasoning do you support?
Harth: At the moment, we can do OWL Lite reasoning without generic transitive properties and cardinality constraints. But we do subClassOf and subPropertyOf. Adding transitivity for arbitrary predicates could be easily added via a set of SPARQL construct queries, and we'll implement that soon.
SWS: Does a comparison of your reasoner with Bossam make sense? Is that kind of technology quite similar?
Harth: There are quite a few differences. Bossam seems to have been tested on data sizes in the order of 100.000 triples. Our dataset from the web has 400 million triples, and YARS2 scales to billion. Bossam seems to have SWRL, general rules, and OWL DL. We focus on a large subset of OWL Lite, but optimise for massive datasets for use in Semantic Web search and large scale data integration. We have a distributed system with combined on-disk/in-memory data structures, and bossam operates in-memory on one machine only. I think the usage scenarios for the two systems are very different.
SWS: Thank you for this interview!
Technical Report
Download Technical Report "YARS2: A FEDERATED REPOSITORY FOR SEARCHING AND QUERYING GRAPH STRUCTURED DATA"