Ontology visualisation and management

[Ian Gordon]

One of the things that we discovered here at Highways England shortly after starting work on our Ontology (see previous post) was that there isn't a huge provision of COTS tools to help you visualise your Ontology, and consequently human readability suffers once you move past a few dozen entities and relationships. Moreover, a key Ontology-building tool (WebProtege) isn't really a commercial offering at all, and as such will probably never be Enterprise-ready (and indeed could go down at any moment!). Other tools such as WebVOWL seem to have paused at the 'nice idea' stage of development, and in general the real development in the market appears to be in the graph database space (Neo4j, Grakn, Stardog) rather than the Ontology-building space. There certainly isn't, as I had hoped, a fully configured GUI where you can filter and dynamically edit your Ontology as you would a mind map without having to resort to form-filling.

If we want our Ontology to function as a schema for a graph database, and that graph database in turn to provide the data storage and logical underpinning of the Digital Twin (and the interface between twins) then limitation to the visualisation and management of our Ontology will consequently limit the functionality and acuity of our Digital Twin.

Key to making this logic work is being able to clearly visualise and edit Ontologies because that will in turn dictate the relationships that it’s possible to model in the Twin. There isn’t really a COTS product that fully fits this brief at the moment, but we have undertaken some research (attached) to assess the market and identify where some investment in products could fill this gap (and indeed wrote some code to make our current tool of choice, Web Protégé, work slightly better). It would make sense to me that if we are seeking alignment in how we create and interface Ontologies, then it makes sense that we also work towards a common toolkit for their creation, maintenance, and visualisation.

The attached PowerPoint describes our market evaluation to date, including where we feel that there is opportunity to augment WebProtege, as well as the obvious opportunity posed by Neo4j Bloom. It seems to me that if we, as a sector, are investing in the creation of Digital Twins, then we should think about how we can work together to build capability in the market to make these tasks easier and more scalable. A common toolkit will reduce our individual investments, whilst maximising the value that we obtain from development.

What do you think?

Ontology_Visualisation_Proof_of_Value_(HE).pdf

[Peter El Hajj]

Hi Ian - thank you very much for sharing this. I am new to data modelling and ontologies, but certainly doing my best to learn. 

And I find both of your posts on DT Hub really helpful, so I look forward to more!   

To comment on your question, having a common toolkit to develop Ontologies makes a lot of sense.  In addition to reducing individual investments,  having a common tool would be useful for knowledge sharing and to grow information management skills across the industry.  It could also facilitate for the ontologies to come together and create one upper ontology for our industry.   

What do others think?

@Neil Thomspon  @Mark Enzer  @RachelJudson

[Andrew Myers]

Hi Ian, 

There is certainly a lot to think about here. Thanks for sharing your ontology, it's great to see how much thought you've put into it.

At Northumbrian Water we're not quite as advanced in our thinking in this area. We've focused on ensuring that we have clear mastering of key data items in our corporate systems and have clear classification and relationships within and betweeen these systems (typically oracle relational databases).  For semi-structured data we've typically used XML or JSON filed in our Data Lake which have a pretty simple ontology and a limited number of links with our more structured databases.

Tools such as the examples above may well have assisted me in my role as Data & Information Architect, but I  doubt many of our end users or other IT teams would be as interested (they just want it to work).

I'd be keen to know who you would expect to use an ontology system, and how you'd be able to keep it in sync with any changes to the data models/classifications of the corporate systems where the data is mastered.

Thanks

Andrew

[Ian Gordon]

Andrew... We've been thinking about the business use of the ontology quite a lot. As you suggested, it's going to seem pretty esoteric to the vast majority of HE staff. Indeed, we had a bit of a running joke within the modelling team that every other day of building the Ontology brought on a existential crisis.

Where I've got to is that the Ontology should act as a force for conformance across the organisation, but may not often be used directly. Rather, most of the time if will exert its influence by providing structure to other artefacts (e.g. logical data models, data catalogues, low level designs etc.).

I find that in HE very often systems are designed without aligning to any business logic. For example CRM systems are designed without clear definitions of 'customer', mutually incompatible network models are created without common definitions. I hope that if we ensure that system design can be traced by to a common set of definitions of the data entities that the organisation manages, then it will be far easier to share and align data both within and outside of HE.

Part of this means being a little circumspect about how much detail to include in the Ontology. If we try to capture and define 1000s of data entities it will be very hard to maintain the Ontology or align with physical systems. But if we aim to capture and maintain the top ~100 data entities, and the crucial relationships sitting between them, then this should be sufficient to guide the organisation without creating crippling bureaucracy! This is also why it's important to have clear hierarchies, not just of entities but also of relationships (e.g. what are the use cases associated with the relationships that we are mapping).

I like the sound of your capturing meta deta in your Data Lake, we've been debating how best to do this ourselves.

[Andrew Myers]

Ian, Thanks for the reply. Your approach makes a lot of sense given the context within your organisation. I think I've been lucky to be appointed Data Architect just as we've been starting a huge overhaul of our core systems, hence have been able to have clear definitions and clear mastering built into these new applications. It hasn't been easy,  but I think it puts us in a great place to better exploit our data for new emerging opportunities such as Digital Twins.

I think documenting our ontology more formally could well be a useful task for me to undertake once we complete our transformation (or when my involvement reduces to free up some time). I expect I'll have similar issues with deciding the appropriate level of detail, so it's great to see your example.

[holgerkessler]

Good morning, new to this group - and hoping to contribute more and more in the coming months....

I would be very interested to stay in touch with both Network Rail and Highways England regarding your underground asset data.

 

[Ian Gordon]

24 minutes ago, holgerkessler said:

Good morning, new to this group - and hoping to contribute more and more in the coming months....

I would be very interested to stay in touch with both Network Rail and Highways England regarding your underground asset data.

 

Welcome!

Do you have any standard specification of information that you believe should be captured for such assets?