Den söndagen den 20:e november 2011 skrev Sébastien HARISPE< harispe.sebast...@gmail.com>: > Hello, > Sorry for the delay in responding, I was busy these days. > > Marko > I will test the proposed stack Thx! > > Mattias > I work on graph based semantic similarity/proximity measures. > In this case the graph is most of the time a reduction of an ontology i.e a > DAG corresponding to the rdfs:SubClassOf backbone of the specification. > Most of proposed measures use set of ancestors/descendants and require a > transitive reduction as preprocess... > I just wanted to known if Neo4J proposed facilities to store temporaly data > during a traversal or if I have to manage them by myself.
The closest would be to begin transaction and modify the graph as you go along to later rollback the transaction. The problem is that you would lock those parts of the graph during that transaction. So it might be better to manage it outside the graph if you'd like to have concurrent traversals over the same subset of the graph > > Thanks a lot > > SH > > On Wed, Nov 16, 2011 at 9:21 AM, Mattias Persson > <matt...@neotechnology.com>wrote: > >> 2011/11/13 Sébastien HARISPE <harispe.sebast...@gmail.com> >> >> > Hi all, >> > >> > First of all, Thx a lot for your amazing work, simply astonishing... >> > Below three questions related to my interest in using Neo4J as a Graph DB >> > on top of which I want to test algorithms dedicated to large semantic >> > graphs (OWL/RDF based). >> > It would be nice if you could help me answer them. >> > >> > >> > A -Using Neo4J with semantic web technologies like RDF/OWL/SPARQL: >> > >> > How does Neo4J interact with these technologies? >> > >> > I found/read: >> > >> > - an old post explaining how to load OWL into Neo4J using Jena API >> > [1]< >> > >> http://sujitpal.blogspot.com/2009/05/using-neo4j-to-load-and-query-owl.html >> > > >> > . >> > - interesting classes in the org.neo4j.meta.model package >> > [2<http://components.neo4j.org/neo4j-meta-model/snapshot/> >> > ,3< >> > >> http://components.neo4j.org/neo4j-meta-model-import/snapshot/apidocs/org/neo4j/meta/input/owl/Owl2GraphDb.html >> > >] >> > but I can't locate the package in the 1.5 apidocs >> > [4]<http://components.neo4j.org/neo4j/1.5/apidocs/> >> > . >> > - that the org.neo4j.rdf and org.neo4j.meta do not follow the core >> > components release, are they currently supported? >> > - a ticket related to the addition of SPARQL plugin to the Neo4J >> server, >> > which means that semantic processes are part of your preoccupations ;). >> > >> > Can you detail the position of Neo4J considering OWL/RDF/SPARQL >> > technologies? >> > I can't easily find documentation about Neo4J and OWL >> > [5<http://wiki.neo4j.org/content/OWL,_SPARQL,_SAIL,_and_RDF>], >> > the use of RDF and SPARQL is well detailed however >> > [6]<http://wiki.neo4j.org/content/RDF_/_SPARQL_Quickstart_Guide> >> > . >> > I also read about Sail on top of Neo4j >> > [7]<https://github.com/datablend/neo4j-sail-test>, >> > and to be honest, I'm quite lost... >> > >> > Do you think Neo4J is an interesting solution to interact with these >> > technologies, that is to say, loading OWL specification as graph >> > (simplification of the spec e.g classes, instances relationships between >> > them), adding RDF triplet and using SPARQL? >> > >> > >> > >> > B - Algorithm implementations: >> > >> > As you know many graph algorithms need to store temporary results e.g. >> > Dijkstra stores shortest path cost for all nodes during execution. >> > Due to the size of the graphs, storing this information in memory is not >> > possible most of the time, so how do you manage this information? >> Temporary >> > node properties? Is there guidance to implement algorithms using Neo4J? >> > >> > For extremely big traversal that could be a problem and the algorithms >> that come with Neo4j (Dijkstra, A*, Shortest path etc.) don't use temporary >> properties. How big traversals are we talking about? I know I've done A* on >> rather many millions of relationships/nodes. Most algorithms touch only a >> fraction of the full graph of course so that's usually not a problem even >> if there are a billion relationships. >> >> > >> > >> > C - In-memory usage of Neo4J: >> > >> > In-memory graphs are very pleasant in order to perform certain CPU >> > consuming algorithmic treatments on a reduced part of the graph. >> > Considering an old post >> > [8]< <http://lists.neo4j.org/pipermail/user/2010-February/002719.html> -- Mattias Persson, [matt...@neotechnology.com] Hacker, Neo Technology www.neotechnology.com _______________________________________________ Neo4j mailing list User@lists.neo4j.org https://lists.neo4j.org/mailman/listinfo/user
