Re: RDB2RDF: a picture is worth a thousand words

Hi Lee

I updated the options and they can be found here [1]

[1] http://userweb.cs.utexas.edu/~jsequeda/rdb2rdf/

I'll go through your comments and then add some more in another email


On Tue, May 4, 2010 at 11:11 AM, Lee Feigenbaum <lee@thefigtrees.net> wrote:

> I don't understand a couple of things in the diagram.
>
> In Option 1, presumably "user specified name substitution" could be far
> more than just name substitution? It could be transformational (via, e.g.
> SPARQL queries or RIF rule sets or what not) to reach the end state
> ontology?
>

Are you suggesting that the local ontology be transformed by the user with
the mapping language? Given the direct mapping and the local ontology, you
would like to change the ontology and therefore also changing the mapping?
If so, are these changes done with our mapping language or with SPARQL
queries/ RIF/etc?

Option 1 is Eric's Isomorphic case and what I'm calling Direct Mapping. The
objective is that the RDF graph is basically equal to the relational schema
(a binary relation becomes a property). So that is why I'm showing only a
User Specified 1:1 name substitution. If there is a change to the graph,
then it is not not isomorphic-direct.

With Option 1, it is straightforward to automatically do the direct mapping,
therefore the labels are generated automatically. Hence the need to do the
1:1 name substitution.



>
> In Option 1, are we saying that the optional stuff on the RHS would be left
> unspecified by this group?
>

No. Optional if a user uses an automatic method and decides to leave the
labels with the automatic ones generated.

>
> In Option 2, why are we starting with "Transform" rather than "Relational
> Schema"?
>
>
This was a typo

> In Option 2, what does the arrow going from "Domain Ontology" up to the
> "transform / non-isomorphic" arrow represent?
>

this was a bad design decision. In the new Option 2, I show the relational
database and how it should be mapped to a domain ontology and how the RDF
graph should look.


>
> Am I correct in understanding that Option 1 is a degenerate case of Option
> 2 in which the transform is obvious and isomorphic and in which "Ontology"
> ends up simply being the local/putative ontology?
>

I'm not sure what you mean here, but hopefully with the new images, it can
answer your questions.






>
>
> On 5/4/2010 11:55 AM, Juan Sequeda wrote:
>
>> We would like to share this and see if we are all on the same page
>>
>>
>> http://docs.google.com/drawings/pub?id=1de31u5wBUheygJTXbxXE7e7H7QI_is9hEXi5dOO5knE&w=960&h=720
>> <
>> http://docs.google.com/drawings/pub?id=1de31u5wBUheygJTXbxXE7e7H7QI_is9hEXi5dOO5knE&w=960&h=720
>> >
>>
>>
>>
>> Juan Sequeda
>> +1-575-SEQ-UEDA
>> www.juansequeda.com <http://www.juansequeda.com>
>>
>