Last year, Hassan Aït-Kaci sent a note to Ontolog Forum (copy below),
and I replied "Your two slide shows are an excellent proposal for
the future development of the Semantic Web." (01)
More recently, he sent me a new report about the updated version
of CEDAR, which is more robust and expressive than last year's.
It is still orders of magnitude faster than popular web reasoners
(Fact++, HermiT, Pellet, TrOWL, RacerPro, SnoRocket): (02)
http://cedar.liris.cnrs.fr/papers/ctr12.pdf
Design and implementation of an efficient Semantic Web reasoner (03)
To evaluate CEDAR, they used six ontologies, for which the number
of sorts (or classes) range from 6,135 for Amphibian to 903,617
for NCBI. (See p.12) (04)
For classification (p. 13), CEDAR and FACT++ are the fastest.
Each one beats the other on 3 of the 6 ontologies. (05)
But after the classification stage, CEDAR is much, much faster than
the others on queries -- by two or three orders of magnitude (p. 14). (06)
Some excerpts from the report: (07)
> For the same queries on the same ontologies, the results achieved
> by CEDAR were compared to those obtained by all the other reasoners.
> The results of experiments show that CEDAR consistently performs
> on a par with the fastest systems for concept classification, and
> several orders of magnitude more efficiently in terms of response
> time for Boolean query answering over attributed concepts, as well
> as for ABox triplestore querying. The latter result is irrespective
> of the triplestore management used because the CEDAR reasoner uses
> its knowledge to optimize SPARQL queries before submitting them
> to the triplestore. (08)
> We use “sort” as a synonym of atomic “class” or “concept.” In
> particular, sorts are partially ordered type symbols, where the
> ordering (“is-a”) denotes set inclusion. (09)
> this version ... can process both TBox and ABox information. It starts
> with a classification phase performed on the TBox which includes cycle
> detection, transitive closure of the “is-a” taxonomic ordering, and
> feature domain/range constraint propagation down the taxonomy, and
> normalization of all such information. Then, the encoded TBox is saved
> on secondary storage independently of any ABox. (010)
> Finally, for retrieving instances, normalized queries are translated
> to SPARQL form that can be optimized using TBox knowledge. In Section
> 3.2, we discuss the classification and the query normalization steps
> and give an example of the SPARQL query generation process. (011)
The above results are for standard RDF triples. But when sort
or type labels are added to the triples, the triple store can be
indexed for even better performance. Depending on the kind of
query, the performance improves by a few more orders of magnitude.
(pp. 16-17). (012)
In many cases, the CEDAR software can use the ontology to analyze
RDF and infer the implicit sorts (or types). But it would be more
convenient for *both* humans and computers to use a notation that
supports typed triples (and n-tuples) explicitly. (013)
John (014)
-------- Original Message --------
Date: Sun, 22 Dec 2013
From: Hassan Aït-Kaci (015)
Hi, (016)
Thought the two presentations below might be of interest to a non-empty
subset of this community. (017)
* http://www.slideshare.net/HassanAitKaci/hak-intis2013 (018)
* http://www.slideshare.net/HassanAitKaci/hak-ontoforum (019)
The 1st one was a keynote at INTIS 2013
(http://intis.conferences-it.com/index.php/tuto-missions/hassanaitkaci),
in beautiful Tangier, Morocco, Nov. 30, 2013 - it is mostly
non-technical, and more forest vs. trees. (020)
The 2nd is more technical and was my (unfortunately hurried)
contribution as part of the Ontolog Forum's panel on Ontology, Rules,
and Logic Programming for Reasoning and Applications (RulesReasoningLP)
mini-series of virtual panel sessions - Nov. 21, 2013
(http://ontolog.cim3.net/cgi-bin/wiki.pl?ConferenceCall_2013_11_21). (021)
Wishing my best Season's Greetings to all! :-) (022)
hak (023)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J (024)
|