Exploring the Role of Ontology in Data Modeling

·May 05, 2022 04:32 PM

Thomas D. Here is a food for thought: having significant data background, I always view the world in terms of data. I consider 'ontology' is a kind of equivalent to 'enterprise data model' - bigger than domain model and involves more than one application area's data needs. However, the representation is in semantic modeling/graph techniques instead of standard ERD or dimensional modeling technique. Any comment?

7 comments

· Sorted by Oldest

Thomas D.
·
Neena D. great comment, a more powerful, multidimensional representation?...I am also interested in what our experts on the channel have to say...also tell us more about your background?
Katariina K.
·
Hi Neena, I share your view of an ontology to be a "kind of equivalent to 'enterprise data model'" capturing central concepts for the business. But it is not only about capturing them but also about giving them an identifier and attaching to them contextual information as well as metadata. That is why it is not only about a modelling technique, but making that information available across systems for reuse. The W3C standards have put tens of yers of RnD to solve and enable this.
👍1
Phil T.
·
Neena D. Sort of… I’ve spent a number of years helping organizations to develop flexible and modular enterprise models, and use both techniques interchangeably (matching the audience). However, it’s more than just the representation style. Traditional enterprise-developed models tend to be overly concrete and brittle. A robust and adaptive ontology often includes or is based on higher levels of abstractions and tends to be more relationship-rich, and here standards like OWL play an important part that you tend not to see expressed in traditional approaches. I suppose a good relatable EDM-type analogy here would be to consider the mind-shift at play that occurs when orgs shift from Kimball/Inmon style models to, say, DataVault…
Neena D.
·
Thomas D. Yes - agreed, certainly multidimensional, It could be more integrated representation when we fold in temporal constraints along with both KG and PG. Even more powerful when conditional logic is mapped in the model since it is outside of ERD or Kimball's dimensional technique currently. Background: trained engineer, practicing data architect - have been in data management field almost three decades. Have seen, done all aspects of data - strategy, governance, metadata, modeling (SQL/noSQL), integration etc.
Neena D.
·
now want to experiment with KG and bridge the gap
Neena D.
·
Katariina K. there are all kinds of IDs captured in the traditional modeling domain. ID - global and internal - for referencing data is a must in modeling world, hence (sometimes) it may become cryptic. Hopefully semantic modeling could alleviate some of those aspects. Thanks for comments!
Neena D.
·
Phil T. Ditto! Couldn't agree with you more. My initial comment was to focus on the approach - many try to get into a new concept area thinking it is completely new territory! Instead, it we just adjust the viewing lens (based on already known domain), it becomes much easier. Who knows I could be wrong??

Thomas D.
·
Neena D. great comment, a more powerful, multidimensional representation?...I am also interested in what our experts on the channel have to say...also tell us more about your background?
Katariina K.
·
Hi Neena, I share your view of an ontology to be a "kind of equivalent to 'enterprise data model'" capturing central concepts for the business. But it is not only about capturing them but also about giving them an identifier and attaching to them contextual information as well as metadata. That is why it is not only about a modelling technique, but making that information available across systems for reuse. The W3C standards have put tens of yers of RnD to solve and enable this.
👍1
Phil T.
·
Neena D. Sort of… I’ve spent a number of years helping organizations to develop flexible and modular enterprise models, and use both techniques interchangeably (matching the audience). However, it’s more than just the representation style. Traditional enterprise-developed models tend to be overly concrete and brittle. A robust and adaptive ontology often includes or is based on higher levels of abstractions and tends to be more relationship-rich, and here standards like OWL play an important part that you tend not to see expressed in traditional approaches. I suppose a good relatable EDM-type analogy here would be to consider the mind-shift at play that occurs when orgs shift from Kimball/Inmon style models to, say, DataVault…
Neena D.
·
Thomas D. Yes - agreed, certainly multidimensional, It could be more integrated representation when we fold in temporal constraints along with both KG and PG. Even more powerful when conditional logic is mapped in the model since it is outside of ERD or Kimball's dimensional technique currently. Background: trained engineer, practicing data architect - have been in data management field almost three decades. Have seen, done all aspects of data - strategy, governance, metadata, modeling (SQL/noSQL), integration etc.
Neena D.
·
now want to experiment with KG and bridge the gap
Neena D.
·
Katariina K. there are all kinds of IDs captured in the traditional modeling domain. ID - global and internal - for referencing data is a must in modeling world, hence (sometimes) it may become cryptic. Hopefully semantic modeling could alleviate some of those aspects. Thanks for comments!
Neena D.
·
Phil T. Ditto! Couldn't agree with you more. My initial comment was to focus on the approach - many try to get into a new concept area thinking it is completely new territory! Instead, it we just adjust the viewing lens (based on already known domain), it becomes much easier. Who knows I could be wrong??