The ORM Foundation

You may want to open a new topic on RDF and OWL in the"Standard" section, next to SBVR. 

Here is a a question I would like to see answered in that section :  How does ORM compare to RDF and OWL ? Among other things, what is the benefit of n-ary predicates compared to binary ones ?  Any pointer to an existing document on the web ?

Thanks in advance !

pcarbonn:

How does ORM compare to RDF and OWL ?

Terry discusses this in the BBB pages 865-871.

Here are some edited extracts:
RDF identifies resources by a URI reference. (ORM identifies object entity types using an identifier)
RDF allows you to state that a class is an instance of itself and so is exposed to logical problems such as Russell's paradox.

OWL does not have a graphic notation. Thus, fact types and their restrictions are declared textually. (Which for me makes it much more verbose)

OWL uses the "open world assumption" . This means that failure to find or to infer some proposition does not imply that the proposition is false.
So it seems to me that OWL is exposed to the logical equivalent of "nulls" which means that in OWL, the answer to the question "Is this proposition true or false" could be either Yes, No or Maybe. (Not very good for logical analyses!)

ORM has the notion of an "atomic fact" which simply put, is an assertion that cannot be expressed in a simpler form without loosing information.
In practice, one sometimes encounters atomic facts which can only be expressed using a ternary, a quaternary or an even higher "n-ary" fact.  
Thus, ORM supports the use of n-ary fact types.
ORM's n-ary fact types can be mapped to binary fact types but with potential loss of information. For example some ORM constraints and derivation rules have no counterpart in the description logics on which RDF and OWL are based.
RDF and OWL are restricted to binary fact types. Thus they are less expressive than ORM.

There are presentations in the ORM Events section of the Library that discuss aspects of your question.
Most of the presentations are based on papers that have been published in Springer-Verlag's Lecture Notes on Computer Science.
Where a paper exists, it is referenced in the description of the presentation.

Hope this helps

Ken 
  

 Thank you Ken.  This is very helpful.

PC

 During my recent visit to the KRDB Research Centre in Bolzano, Italy, Enrico Franconi introduced me to a book called "The Description Logic Handbook.  After a little more research, I found another book called "Description Logic Framework to Synchronize ORM Model and OWL Language" - published in 2012.

I hope you find these books useful.

Ken

 I have a follow-up question.   I was recently looking at the Visual Owl standard which attempts to visually represent an ontology for more casual users who do not know or regularly deal with database or software design notations.   Visual OWL is a proposed graphical standard for OWL.  The Visual Owl link is at: http://vowl.visualdataweb.org/  and contains some some nice overview papers.   

 1.  Is Visual OWL a good representation of  RDF/OWL and is it easy to use?

2.  Are there facts I can model in ORM that I cannot model in Visual OWL? 

3.  Assuming I accept the Open World Assumption, the limitation to binary facts-types, the violation of Russell's Paradox and the limitations of Literals to only the Object part of a Triple is Visual OWL a "better" notation than ORM?   Should I use Visual OWL and not ORM when constructing ontologies? 

    3a.   Is Visual OWL better for casual users and then use ORM for designers?  or

     3b.  Given that a SME can more easily understand Visual OWL and it is executable (generates OWL/RDF) why not use Visual OWL in all the Requirements, Design, Implementation, and Production phases? 

4.  How is is ORM or Visual OWL in setting up data analytic solutions (e.g., evaluate the credit risk of different kinds of customers)?

 5.  Based on your analysis of UML, ERDs, and ORM, how do you think a visual language like VOWL would fare?

 Finally, it it possible to get a copy of research at Liverpool on UML, ERDs and ORM?

Thanks,

Alan Blair 

 Hi Alan,

Thanks for your questions.

I have never used Visual OWL so I will need time to look into it.

Regarding question 5, my University of Liverpool MSc Dissertation has never been published.
However, I plan to include it with the book that I am writing - which I estimate will be published in mid 2016.

The dissertation document itself  is 130 pages.

I used the scientific method of making hypotheses and then devising quantifiable experiments to test their truth value (True or False)
One of the results that I found to be most surprising was about the relationship between the amount of a person's formal training in UML and the amount of time the person needed to create the UML class model from the 331 words of text.

My hypothetical prediction was that people with formal training in UML would take less time than those who are self taught.
From this, I devised and tested two hypotheses:
H0: There is no difference in the effort (time) required by those with formal training and those who are self-taught.
H1: Those with formal training will require less effort to create each UML class model.

The experimental evidence showed that H0 is true and H1 is false.

Ken  

Hi,

1. No, for many reasons. VOWL ignores most of the OWL contracts, presenting only an abstract (and sometimes illogical) summary of the conceptual model. It does not scale up when conceptual models have more than, say, 10 classes. It can not be used to design conceptual models but only to visualise summaries of existing conceptual models. 

2. ORM is strictly more expressive than OWL2 anyway. VOWL presents only some aspects of an OWL conceptual model.

3. No. Because of the above, but also when one compares ORM with OWL. Indeed, I claim that ORM is already an (almost) ideal graphical representation of arbitrary OWL conceptual models, and it provides excellent design tools and most importantly design methodologies.

3b. Ideally, you may want to generate OWL from an ORM schema. We are working on a well founded and also practical solution to that, by enriching the NORMA tool with a "smart" "export" function to OWL.

5. ORM is the best!

cheers

--e.

 Enrico Franconi                  - franconi@inf.unibz.it

 Free University of Bozen-Bolzano - http://www.inf.unibz.it/~franconi/

 Faculty of Computer Science      - Phone: (+39) 0471-016-120

 I-39100 Bozen-Bolzano BZ, Italy  - Fax:   (+39) 0471-016-009

Thanks for the quick responses.  My heart is certainly with ORM.    I ran across it around 2002 and fell in love.   In most of my projects I model in ORM and then switch to whatever notation the project is using.  I can see my mistakes more clearly in ORM.  When no notation is mandated I just use ORM.  Also, I find it a great notation to validate my models with SMEs.  

I am working on a project with a large financial services firm.  They  Enterprise Management Group wants to create a customer ontology (a toy to start).  They are looking at ERDs, ORM, and UML.   Another data group is using Hadoop/Hbase to access and analyze data warehouse data sets.  They propose that an analyst model that data he wants to use as an ontology.  This tool is not implemented yet, but they are very interested in using VOWL as a visual language to represent these datasets.    Both groups are talking about using a common notation so that Data Modelers, Implementers and Data Analysts can share a common visual language (a beautiful dream).    An ideal would be to use this notation to also represent APIs.  Over the next seven weeks we will be doing an evaluation of the pros and cons.  I have read 1.6 of the VOWL papers and am starting to work out how to approach this comparison.   Hence, my interest in Ken's research.  The VOWL folks claim to have done a some usability tests so I will review them are well.   So far I have not found a VOWL to ORM comparison.

 Could you please be more specific as to why:

1.  VOWL ignores most OWL2 constructs (which constructs?) and is only an abstract and sometimes illogical summary of a conceptual model?    In the examples, I have seen I noticed that they implicitly re-introduce attributes as members of a larger object.

2.  Why VOWL has scaling problems (e.g., 10 concepts) that are worst than ORM?

3.  Why VOWL can visualize a summary of conceptual model, but cannot design a conceptual model (no associated methodology?)? 

4.  How ORM is more expressive than OWL2  and in turn is OWL2 is more expressive than VOWL.

5.  Do these difference in expression have pragmatic impacts (e.g., lead to significant modelling errors) or are the just fine points that logicians would argue about?

Finally, is there a target date for when this  well founded and also practical solution of exporting OWL from the NORMA tool will be released?    I noticed that NORMA can export to a number of other notations, but not yet to OWL.  I recall Dr. Halphin talking about adding an OWL export back in 2011 and have read most of his OWL to ORM comparison papers.  I agree this is the ideal approach, but is this smart export of RDF/OWL2 something that is going to be released within the next five months?

To my mind ORM has had 30 years of people beating up this notation and improving it; it should be a much more stable notation and able to represent more domains of discourse and share them among stakeholders.   But I need to make this argument in a lot more detail and then construct a concise summary.

Thanks,

Alan

Hi Alan

For some further details on how ORM compares with OWL and RDF, please see the fifteen articles I wrote on Ontological Modeling for the Business Rules Journal. You can download these from the Resources section of my website www.orm.net.

Cheers

Terry 

As a user of ORM as well as OWL and numerous databases.  I view ORM as the analysis tool of choice and OWL and databases as deployment options for expressing the logical structures that are captured and verified in ORM.  Each deployment option has features and drawbacks and the choice depends, at least in part, on the purpose of the project. 

The lack of nary objects in OWL can be compensated for with reification. Sometimes this is straight forward as in the case of a fact such as, "John assigned Bill to design Project 1."  Where John, Bill and Project 1 are elements of the key.  Owl could be made to handle this by introducing the concept of Assignment and creating three facts.  Assignment 1 was made by John.  Assignment 1 was given to Bill.  Assignment 1 was for Project 1.  Because Assignment is a concept most people are familiar with this is only a little unnatural.  If the overarching concept is one that does not have a convenient name the discussion with the SMEs becomes more difficult and going to the ORM model makes verification easier.

The open world assumption is also  related to the purpose of the system being constructed.  For systems that manage a closed world such as the employees of a company, the open world assumption is difficult to work with since it leaves us not knowing who else might be working at the company and requires listing all of the people who don't work there as explicit facts to close the world.  That is problematic at best.  However, for a system that is trying to assemble knowledge about an evolving research area, the open world assumption is quite appropriate, because not all facts are not yet known and sometimes the correct answer to a query is maybe since current knowledge can not answer the question.

I hope this adds to the discussion above.

Jack