Best Practices for Modeling RDF and Visualizing Relationships

·Jun 26, 2024 03:18 PM

I'm starting to learn RDF at the same time I'm learning OntoText, and I'm struggling:

1.
What is the best syntax for RDF to use to model things?
2.
How elaborate do I need to go beyond identifying entities and their relationships to get a nice visual of the things?

One of the challenges I'm having is just "being able to see all of my objects and their relationships" without all the data types or that persistent "CLASS" type polluting the visual. I suspect that is some SparQL magic required, but it feels like a huge lift just to get a nice picture of all the things that I have in a logical way.

8 comments

· Sorted by Oldest

Matthias S.
·
Hi Robert, A few clarification before so you can make sure RDF is the right fit for your need:
RDF is a data exchange format (representing everything as a set of triples (s, p, o), which all together forms a graph.
Then there are some languages that defines some particular triples to which some meaning are associated, i.e., a vocabulary a community is agreeing on when sharing data. RDFS and OWL are some examples.
You can visualise a set of RDF triples in various ways such as text files in different serialisation format, e.g., Turtle or RDF/XML. Or you can use a great tool like OntoText which reads your RDF triples to show you a nice graph.
Usually when modelling with RDF people tend to think that what they have defined in RDF as an RDFS or OWL class is an entity and therefore should appear as a node in the visual graph. However, what you really do in this cas is inserting the triple (myThing rdf:type rdfs:Class). And if I remember well the graph visualizer from OntoText by default shows the raw rdf triples. It also often happened that some basic RDF triples are preloaded when creating a triplestore (e.g., OWL or RDFS triples). Hence, the default graph vis will show them. With this lengthy response I hope to clarify that if your only goal is to visualise your data as a graph, RDF might be a complex choice. That being said, OntoText graph vis lets you define the SPARQL queries to execute for each "click" action on do when manipulating the graph vis. I am telling you all this from my little previous experience with OntoText and my understanding. I hope it helps and I am very happy to discuss. Good luck 🙂
Wolfgang S.
·
Sometimes some additional tooling helps. We at metaphacts provide comprehensive tooling for Knowledge Graph Management and Modeling and have good support for OntoText‘s GraphDB. There is a free trial available that you can run on your own system using Docker which may help you with loading, exploring, visualizing, searching and modeling RDF data. See this blog post for an example: https://blog.metaphacts.com/generating-value-from-your-knowledge-graph-in-days
Robert H.
·
Thanks! This is a precursor to trying to use knowledge graphs with AI to have those key entities and relationships available to validate incoming, unstructured information and create new ideas. It seemed like "RDF" was the basic means of expressing that knowledge, and so I set about trying to find a proper tool to visualize and create whatever constitutes a model in this case that AI can use. And then I can visualize with. I don't know if that helps you understand my struggle. I'm a former data modeler, so maybe those old habits are dying hard. Also used to do a lot of concept mapping, nothing in a formal or automated way.
Matthias S.
·
Ah! Then in my humble opinion, you are on the right path 🙂 I think I understand your struggle with the visualisation. You might want to make sure you understand what are the triples really present in my triplesotre (often it is more than simply the ones you inputed) and what are the triples I really want to show. Then build the corresponding SPARQL queries to finetune your visualisation. If you have a sample data, I can try carving out some time to test some vis (I can not promess I will have the time though 😉 )
Robert H.
·
It's fine! I appreciate it, and could use a link or reference to a primer on RDF that's not dry; I read a lot that are about the nuances of the file format and not how to go about modeling something. So I get tied up in instances and properties and often end up with a giant "CLASS" in the middle of the diagram that everything connects to, which isn't useful. I know things are derived from or of type class
Bob DuCharme
·
Rob H. You can also learn a lot about the relevant standards at https://lms-academy.ontotext.com/, which is free. (Full disclosure: I am an Ontotext employee.)
Robert H.
·
Thanks for the disclosure, and the link!
Robert H.
·
I have to say that I'm utterly flabbergasted to have had a call with Ontotext today pre-sales that was Simple awesome. Answered my questions, taught me some things I did not know about RDF modeling, and were just incredible. Felt like I was talking to mentors and helpers and not "salespeople." Good job, Ontotext. I look forward to the next call.

Matthias S.
·
Hi Robert, A few clarification before so you can make sure RDF is the right fit for your need:
RDF is a data exchange format (representing everything as a set of triples (s, p, o), which all together forms a graph.
Then there are some languages that defines some particular triples to which some meaning are associated, i.e., a vocabulary a community is agreeing on when sharing data. RDFS and OWL are some examples.
You can visualise a set of RDF triples in various ways such as text files in different serialisation format, e.g., Turtle or RDF/XML. Or you can use a great tool like OntoText which reads your RDF triples to show you a nice graph.
Usually when modelling with RDF people tend to think that what they have defined in RDF as an RDFS or OWL class is an entity and therefore should appear as a node in the visual graph. However, what you really do in this cas is inserting the triple (myThing rdf:type rdfs:Class). And if I remember well the graph visualizer from OntoText by default shows the raw rdf triples. It also often happened that some basic RDF triples are preloaded when creating a triplestore (e.g., OWL or RDFS triples). Hence, the default graph vis will show them. With this lengthy response I hope to clarify that if your only goal is to visualise your data as a graph, RDF might be a complex choice. That being said, OntoText graph vis lets you define the SPARQL queries to execute for each "click" action on do when manipulating the graph vis. I am telling you all this from my little previous experience with OntoText and my understanding. I hope it helps and I am very happy to discuss. Good luck 🙂
Wolfgang S.
·
Sometimes some additional tooling helps. We at metaphacts provide comprehensive tooling for Knowledge Graph Management and Modeling and have good support for OntoText‘s GraphDB. There is a free trial available that you can run on your own system using Docker which may help you with loading, exploring, visualizing, searching and modeling RDF data. See this blog post for an example: https://blog.metaphacts.com/generating-value-from-your-knowledge-graph-in-days
Robert H.
·
Thanks! This is a precursor to trying to use knowledge graphs with AI to have those key entities and relationships available to validate incoming, unstructured information and create new ideas. It seemed like "RDF" was the basic means of expressing that knowledge, and so I set about trying to find a proper tool to visualize and create whatever constitutes a model in this case that AI can use. And then I can visualize with. I don't know if that helps you understand my struggle. I'm a former data modeler, so maybe those old habits are dying hard. Also used to do a lot of concept mapping, nothing in a formal or automated way.
Matthias S.
·
Ah! Then in my humble opinion, you are on the right path 🙂 I think I understand your struggle with the visualisation. You might want to make sure you understand what are the triples really present in my triplesotre (often it is more than simply the ones you inputed) and what are the triples I really want to show. Then build the corresponding SPARQL queries to finetune your visualisation. If you have a sample data, I can try carving out some time to test some vis (I can not promess I will have the time though 😉 )
Robert H.
·
It's fine! I appreciate it, and could use a link or reference to a primer on RDF that's not dry; I read a lot that are about the nuances of the file format and not how to go about modeling something. So I get tied up in instances and properties and often end up with a giant "CLASS" in the middle of the diagram that everything connects to, which isn't useful. I know things are derived from or of type class
Bob DuCharme
·
Rob H. You can also learn a lot about the relevant standards at https://lms-academy.ontotext.com/, which is free. (Full disclosure: I am an Ontotext employee.)
Robert H.
·
Thanks for the disclosure, and the link!
Robert H.
·
I have to say that I'm utterly flabbergasted to have had a call with Ontotext today pre-sales that was Simple awesome. Answered my questions, taught me some things I did not know about RDF modeling, and were just incredible. Felt like I was talking to mentors and helpers and not "salespeople." Good job, Ontotext. I look forward to the next call.