|
|
|
| e-Pub |
Previous | 
Home
Section: New Results
Knowledge Graphs
SPARQL Template Transformation Language
Participants : Olivier Corby, Catherine Faron-Zucker Song.
We finalized the design and implementation of SPARQL Template Transformation Language (http://ns.inria.fr/sparql-template/) (STTL) [58] , [73] . STTL is an extension of SPARQL with a template {} where {} clause that enables a transformation to generate presentation format for RDF. For example, it is possible to generate Turtle, OWL or SPIN-to-SPARQL syntax.
We designed a new service in Corese server that returns HTML. Using STTL transformations that generate HTML, we are able to set up light weight Semantic Web servers on top of local RDF Datasets or remote Datasets such as DBpedia (http://corese.inria.fr ).
RDF Serialization and Introspection
Participant : Olivier Corby.
We started a work on RDF serialization of (Java) objects for Semantic Web system introspection. In conjunction with the overloading of SPARQL named graph pattern, we are able to query the system on several internal status such as graph index, triple provenance, property path triples, etc.
OWL 2 RL
Participant : Olivier Corby.
We dramatically optimized Corese Inference Rule Engine and we were able to run OWL 2 RL (http://www.w3.org/TR/owl2-profiles/#OWL_2_RL) rule base on the FMA ontology (Fundational Model of Anatomy (http://sig.biostr.washington.edu/projects/fma/release/v3.2.1/alt_formats.html)) with interesting performance. The initial OWL graph contains 1.74 million triples, the final graph contains 13.46 million triples and the rule engine runs in 3 minutes.
Rules for the Web of Data
Participants : Oumy Seye, Olivier Corby, Catherine Faron Zucker, Fabien Gandon.
This year we focused on the validation, the update of rules bases and the optimization of the reasonning. The goal of this work is to detect some inconsistencies in selected rule bases with respect to ontology and offer users to correct this. We built a set of SPARQL queries enabling (1) to build specific rule bases for a given context or application, (2) to optimize inference engines based on rule selection with respect to target RDF data sources, (3) to validate and update rule bases. We propose another optimization of the inference engines based on graph of rules dependencies and rules aplication ordering. This work is published in [62] .
KGRAM
Participants : Olivier Corby, Fuqi Song.
We received a two years grant from Inria to support the development of the Corese platform. This action aims at enhancing Corese software for conforming to latest W3C standards and facilitating its usage in distributed environment, we integrated several open source parsers to Corese, including JSON-LD, RDFa, TriG and N-Quads. Corese now is able to process RDF dataset in these formats. A Firefox extension called RDF Triple Collector (RTC) was developed, it can extract triple from web pages (annotated using RDFa), upload triples to Corese server and query data using SPARQL endpoint. A prototype of LDP 1.1 (Linked Data Platform) is implemented using RTC as data collector.
Besides, with the purpose of improving Corese query performance and carrying out research work on distributed environment, we proposed and developed a heuristic-based query planning method within Corese. The approach includes 3 main steps: 1) generate extended SPARQL query triple pattern Graph (ESG), 2) estimate the cost of ESG using pre-defined heuristics and cost models and 3) search ESG to find a good query plan and rewrite the SPARQL query. The approach was evaluated using BSBM benchmark, the results suggest that the developed method optimized 60% of the query execution time averagely [77] .