- From: Aliabbas Petiwala <aliabbasjp@gmail.com>
- Date: Tue, 21 Jun 2011 17:12:10 +0530
- To: public-lod@w3.org
Is neo4j http://neo4j.org a good option to consider for this ? does it provide seamless integration with dbpedia, free base etc? On 6/19/11, Marco Brandizi <brandizi@ebi.ac.uk> wrote: > Hi Aliabbas, > > It all depends on what you want to represent and which tasks you want to > perform. > OWL-based models (and RDF/OWL triplestores) are useful when you have > heterogeneous data, for which you cannot define a schema a-priori (eg, > http://dc-research.eu/, OBI ontology). > OWL modelling can add useful inference capabilities to traditional OO or > relational modelling (eg, find automatically that the brother of my dad > is my uncle). > RDF/OWL world are particularly interesting if you want to integrate your > data with similar knowledge, which is increasingly being exported via > http and via RDF/OWL (ie, linked data, see eg, DBpedia, Freebase). > Relational DBs are useful to back object models (so, schema-based > models), you may want to consider NoSQL and Document-based stores too. > Exporting OWL from existing DB may make sense either, especially when > you have well-defined entities (ie, a schema), that you annotate with > ontologies or alike (eg, biomedical records, annotated with biomedical > ontologies, like MESH or GeneOntology). > > Also, keep in mind that in the case of OWL modelling you will typically > need to change your mindset completely. While a relational (or OO) > database is a closed dataset, which defines all (and only) the known > facts extesively, an OWL-based knowledge base is an open knowledge base, > where only what is known is defined by means of logical statements, > keeping the unknown as unknown, not as "not true", or "not existing". > For instance, in OWL it's difficult to establish that person1 and > person2 are two different individuals, cause usually nothing excludes > that they are the same real thing. Most straightforward way is to > explicitly say that they are owl:differentFrom (but, you need to do it > for all the pairs) and this is obviously different than a relational DB, > where two tuples are usually considered different by default (OWL-2 has > introduced identifier properties, but not sure they work so great). As > another example, it's difficult to use OWL for enforcing the equivalent > of relational integrity. For example, you may say that an employee must > have one boss, but this is an axiom, something that can be used to infer > that employee1 has such boss and we can call him/her emp1Boss. You > cannot say your KB is inconsistent just because you cannot find an > explicit declaration of a boss for an employee in that KB, ie, it may > have been declared somewhere else (or nowhere, but it still holds true > in reality). People often make confusion on this sort of things. For > instance, when one says that every protein has a standard international > name (it happens they do that in OWL), they actually mean that every > identified and classified protein that has a record in some > international biological database, has also an associated name in that > DB. That doesn't exclude proteins that exist, even if we know them and > we haven't named them yet. This comes from the fact that "has-name" is > not an inherent property of a protein, in the same sense "has-father" is > for a living being, since the former is just a conventional requirement > for the representation of the real thing. > > Hope it helps. > > Marco. > > > > On 18/06/2011 08:04, Aliabbas Petiwala wrote: >> we are looking forward to make an OWL ontology database as a mirror of >> a relational database for an upcoming social network semantic website >> is it a good decision to make? >> >> actually we can go for three options: >> >> 1. complete owl database, no relational db >> 2. owl mirror of relational db >> 3. only relational >> >> we are very much interested in the second option is it wise to use >> mirror of ontology database as a relation database? how can owL >> Ontology database be efficient than a relational one considering that >> we will need to query a lot of external databases like >> dbpedia,freebase etc? >> >> for the second option, ontological databases requires us to make a >> query against hundreds of different schemata and classes properties >> which seems to be a costly affair. And is it a wise decision to go for >> a complete ontological database for a social networking website ? The >> project is involves sharing of lot of small chunks of information >> across an array of distributed users building a personalized model of >> the user. >> >> ontotext, http://www.systap.com,http://www.opencalais.com/ provides a >> solution for semantic repositories , its really difficult to determine >> which is the best option for a STARTUP like us and we can't go in for >> these expensive paid consultations. so what are the options and >> solutions availaible for us? >> >> > > -- > > =============================================================================== > Marco Brandizi<brandizi@ebi.ac.uk> > > Microarray Group - Sr Software Engineer > http://www.ebi.ac.uk/microarray > > European Bioinformatics Institute > Hinxton, CB10 1SD, United Kingdom > Office A3141 > > Tel.: +44 (0)1223 492 613 > Fax: +44 (0)1223 492 620 > > http://www.marcobrandizi.info > > -- Aliabbas Petiwala M.Tech CSE -- Aliabbas Petiwala M.Tech CSE
Received on Tuesday, 21 June 2011 11:42:40 UTC