W3C home > Mailing lists > Public > public-xg-rdb2rdf@w3.org > November 2008

Re: CORRECTION Terminology: Data Source & Domain Ontologies

From: Juan Sequeda <juanfederico@gmail.com>
Date: Tue, 25 Nov 2008 16:09:58 -0600
Message-ID: <f914914c0811251409y205c1b71g5b6c7592e9718705@mail.gmail.com>
To: "Kingsley Idehen" <kidehen@openlinksw.com>
Cc: "Satya Sahoo" <sahoo.2@wright.edu>, public-xg-rdb2rdf@w3.org
I see two aspects.

If the SQL schema has been modeled with rich semantics, then applying a
direct mapping method would automatically create the ontology based on the
sql schema, hence automatically create RDF from the RDB. So this is an easy
step for who ever needs a quick and dirty solution. I just submitted a paper
to WWW09 where we show how direct mapping can be enough and useful.

The other aspect is that by creating the data source ontology, the problem
has been shifted to another representation: an ontology to ontology mapping
problem. The semantic web depends also on mapping ontologies, so this is
something that I believe will be solved little by little.

The poster that I presented at ISWC elaborates on my vision of solving the
integration of relational databases and the semantic web in a 2 step
process. First step is to do direct mapping. Second step is to do ontology
matching incorporating other techniques. See here [1] for more detail if


Juan Sequeda, Ph.D Student

Research Assistant
Dept. of Computer Sciences
The University of Texas at Austin


Semantic Web in Austin: http://juansequeda.blogspot.com/

On Mon, Nov 24, 2008 at 9:12 PM, Kingsley Idehen <kidehen@openlinksw.com>wrote:

> Satya Sahoo wrote:
>> In the spirit of moving towards RDB2RDF lexicon, as I understand mapping
>> of RDB to RDF can be viewed from two perspectives (using DL terms):
>> 1. RDF data (ABox) generated without the creating or using an existing
>> model/ontology schema (TBox) - The mapping logic is simply captured in code
>> or as expressions (often XPath etc.).
>> 2. RDF data (ABox) + a model/ontology (TBox - often in RDFS/OWL) - In this
>> case, I believe the "Data Source ontology" is a bootstrapping step that
>> exploits the mappings rules from Tim Berners Lee (1998),"Relational
>> Databases on the Semantic Web",  "Table -> Class" and "Column -> Property"
>> with additional refinements (for FK etc.).
> "Data Source Ontology" is a simple TBox derived from the SQL Schema based
> on:
> Table-->Class
> Column->Attribute (one kind of property)
> Foreign Key-->Association/Relationship with another Class (another kind of
> property)
> Relational Databases on the Semantic Web re. the mapping above is basically
> what NeXT offered with EOF (with current day rudiments in Mac OS X Core Data
> Services).
> The same approach applies to .NET's Entity Frameworks.
> My key point here is that the concept of mapping RDBMS schemas to Entity
> Models pre-dates the Semantic Web.
>>  For real world applications, as we see in biomedical domain or ordnance
>> survey, incorporating domain semantics is essential, preferably in form an
>> explicit "Domain ontology", for the RDF instance to be of any practical use.
>> A "Domain ontology" may either be created "Top-down" by domain experts from
>> scratch - such as the biomedical ontologies listed at the National Center
>> for Biomedical Ontologies (NCBO) - Open Biomedical Ontologies (OBO), or
>> "Bottom-up" where "Data Source ontologies" are used as initial input and
>> then enhanced further by domain experts.
> Yes.
>>  But, once a "Domain ontology" has been created, I don't see the need for
>> "Data Source ontology", since in my view "Domain ontology" will be
>> "superset" of "Data Source ontology" and applications can directly interface
>> with the RDB or create a "RDF dump" (often called ontology population) -
>> using the mapping logic  in "Domain ontology". If the "Data Source ontology"
>> satisfies the requirements of an application then also we have a single
>> ontology - the "Data Source ontology" (= "Domain ontology"). I am not sure I
>> see the need/benefit for multiple-levels of ontologies and bring in mapping
>> issues between them, which in my view is a redundant exercise.
> In some cases you can just generate a basic TBox for an RDBMS schema (
> "Data Source Ontology") and then focus the rest of the mappings to effort en
> route to a  "Domain Ontology" without touching the ABox. Of course, in some
> cases you can map directly to the "Domain Ontology", but it shouldn't
> deemphasize the utility of the "Data Source Ontology". Thus, I don't see any
> mutual exclusivity here, each approach offers value, and for continuity re.
> knowledge development and dispatch re. RDB2RDF, the "Data Source Ontology"
> is a great conduit (since this is the common pattern already in play outside
> the RDF realm).
> Kingsley
>>  Satya Sahoo
>> http://knoesis.wright.edu/researchers/satya
> --
> Regards,
> Kingsley Idehen       Weblog: http://www.openlinksw.com/blog/~kidehen<http://www.openlinksw.com/blog/%7Ekidehen>
> President & CEO OpenLink Software     Web: http://www.openlinksw.com
Received on Tuesday, 25 November 2008 22:10:39 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 19:39:03 UTC