John and all,
I also think that building an ontology 'bottom up', from the logical
and relational schema
abstracting the representation of conceptual schema, is an interesting
methodological option. (01)
In a way, it's a kind of reverse engineering approach, that would
produce an exact ontological representation of an existing 'system'. (02)
We often use this techinque implicitly, and we call it 'mapping'
(knowledge, data, metadata, schema mapping). (03)
However, in order to build a relational schema we need a conceptual
schema first, and some form of ontology needs to support the
conceptual schema development
So in the development lifecycle terms, we need the product of what we
are trying to achieve (the conceptuaization) before it can be
generated by the associated relational schema that is derived from it.
How do we go about that? (04)
This problem could perhaps be overcome by undertaking the schema
development iteratively, which means have a notional, semantic,
conceptual schema first, developed based on a first analysis, derive
the logical and relational schemas from that, then abstract the exact
map of the representation of the ontology from the relational schema,
as suggested in the first post. (05)
If we chose this method however, we should be aware of its limitation.
A logical and relational schemas are often the result of some
modelling exercise, and of some compromises that we make at system
design level to 'get the job done'
It's surely not a complete nor true representation of reality, but of
a model to fit a precise purpose andor function.
An ontology derived from a conceptual schema can only be a 'partial'
knowledge representation, ie the ontology of a 'closed' system, and
never the ontology of a domain, of an open system or of an upper
So I would think that this method is suitable mainly for functional
suited only to some cases, but surely it should be in the 'manual of
ontology develpment' (07)
Paola Di Maio (09)
On 3/8/07, John F. Sowa <sowa@xxxxxxxxxxx> wrote:
> Juan and Patrick,
> That is an excellent thing to do:
> > I was wondering if anybody had information or comments
> > about generating ontologies using the schema of a
> > relational database, not a mapping of a DB and an
> > existing ontology, but creating a ontology from scratch
> > using only the DB schema.
> In fact, many people in the database community recognized
> the strong correspondences between the issues and techniques
> of database design and knowledge base design very early.
> For example, my first published paper on conceptual graphs
> (in 1976) explicitly addressed the problem of mapping
> natural languages to database languages. And I used the
> term "conceptual schema" for certain kinds of CGs before
> I learned that the DB people were also using that term:
> Conceptual Graphs for a Database Interface
> Some people thought that databases were my major inspiration
> for CGs. But in fact, I started with NL semantics. Since I
> was working for IBM, I realized that it was important for me
> to be "relevant" by relating what I wanted to do to what IBM
> thought it would be useful for me to do.
> When I wrote my _Conceptual Structures_ book in 1983, I
> related the work in natural languages, knowledge bases,
> and databases to the many millennia of research in
> philosophy. In fact, I was one of the early adopters
> and proponents of the term "ontology" in that book.
> (If I had known then what I know now, I probably would
> have chosen a different word.)
> In any case, to answer your question, the people who had
> been organizing the annual VLDB (Very Large DataBase)
> conferences for many years devoted an enormous amount of
> attention to doing and redoing exactly the same kinds of
> things that people are doing today.
> The DB people got there first by addressing the problems
> of "database alignment" in the 1970s. The SRKB people
> did the same in the 1990s. The SemWebbers are doing
> the same today.
> Every 20 years or so, a new generation of people comes
> along and reinvents the same wheels, thinking that nobody
> else had ever done anything before them.
> For more on the history of some of these ideas, see the
> slides of a talk I gave last year:
> Extending Semantic Interoperability.
> And to comment on Patrick's note,
> PD> While certainly useful, logic should not be portrayed as
> > the warp and woof of ontologies or ontological work.
> I agree. Ontology and logic are different subjects. But I
> would add that *every* precise declarative notation that is
> capable of making statements that are judged true or false
> *is* a version of logic.
> UML diagrams, for example, are a version of logic. The
> declarative subset of natural languages forms a version of
> logic -- and in fact, NL (actually Greek) was Aristotle's
> inspiration for the first formal logic. And *every*
> statement in *every* version of logic that anyone has
> ever invented can be translated into natural language.
> So anyone who is doing any work on ontology is stating
> their results in some version of logic (but possibly in
> a rather imprecise and insufficiently understood version).
Paola Di Maio
School of IT, MFU.ac.th
"For as long as space and time endures
may I too abide to dispel misery and ignorance" (013)
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (014)