Debunking some RDF-vs-Property Graph Alternative Facts - Dr. Jesús Barrasa, Neo4j

[@32to28]

Sort of misleading how all RDF graphs are presented as relational databases at 12:00, which sets up the strawman argument that there are a lot of joins. I'd sort of expect something better from neo, but hey. Marketing is marketing.

[@jbarrasa4649]

Hi @32to28, thanks for your comment. How are RDF graphs persisted in your opinion if I may ask? Are all triple stores not strongly index-based (and some even backed by relational DBs)? Is triple navigation not based on the principle of a join operation? Maybe you know of a triple store that takes a different approach and want to share?

[@MrLonelyOffc]

It's really great sir. I clearly understand the difference of RDF and Neo4j. I, even, saw your blog posts and other videos. I really like to use Neo4j in my NLP related project.

[@donha475]

Super helpful thanks... I'm new to this area... really helped me to plan my little project...

[@keithvaneaton5840]

This is very helpful, especially the demo around 15:00. Question: What specific plugin(s) are you using to call the stored procedure of semantics.importRDF?

[@jbarrasa4649]

neo4j.com/labs/neosemantics/

[@subhashisdas81]

It is very smart and like a marketing presentation while skipping some SoA or comparison with GraphDB such as ontotext graphDB. LPG is not a W3C recommendation where RDF is which makes a big difference in the open community.

[@jbarrasa4649]

Hi @subhashisdas81 as I'm sure you know, after reading the W3C recommendation you mention, RDF is a model for data exchange. In that recommendation you will find absolutely no mention of how to store or manipulate your data. You can persist your graph as a property graph in Neo4j and still be 100% RDF compliant when you expose your data for external consumers (same when you import it). It's not marketing, it's factual. Also, to your point about caring for the open community, here's the link to Neo4j's codebase on GitHub (github.com/neo4j/neo4j). Could you point me to the same for Ontotext GraphDB?

[@hodaslimani2945]

Great thank you!

[@markhall7173]

Let’s say you’re creating an enterprise ontology to describe business processes. Doesn’t Property Graph approach risk ‘properties’ being having to be defined at each object level? Although he RDF approach creates a denser graph, it does seem to force at least some standardization of property relationships. Triples can be useful that way and domain ontologies more consistent. I understand consistency and governance might not be the primary use-case for the LPG approach, but within enterprise it’s often useful to have such constraints and structure to drive common approaches.

[@jbarrasa4649]

Hi Mark, thanks for your comment. It is perfectly possible to define an enterprise ontology as an LPG, there are multiple cases of use of Neo4j for data governance. OWL and RDFS are just one way of formalising vocabularies/ontologies. I could agree (if that's what you mean) that RDFS/OWL have well-defined semantics that can be used for inferencing or consistency checking but again, there are ways to achieve the same using LPGs (if you're interested have a look at jesusbarrasa.wordpress.com/2016/04/06/building-a-semantic-graph-in-neo4j/). I'd be curious to know what RDFS/OWL primitives you use in your enterprise ontology definitions and whether these include anything beyond classes and properties definitions. Happy to continue the conversation here, but you can also join our slack channel neo4j.com/developer/slack/

[@siddharthdedhia11]

@@jbarrasa4649 Hi Dr. Jesús , It might be a stupid doubt but the reasoning capabilities of neosemantics are very limited. Are there any reasoners existing that do not use ontologies/formal vocabulary that are comaptible with neo4j?

[@jbarrasa4649]

Hi @markhall7173 No it does not. Here's how you create a graph in neo4j using an ontology (kzbin.info/www/bejne/qJXIZJShlK1qZrs, kzbin.info/www/bejne/Zma6nJpnpZhnm80). Also, FYI, ontologies DO NOT enforce any kind of constraints on an RDF graph. Ontologies are purely descriptive and useful for inference. x.com/BarrasaDV/status/1303662848422424577 Consistency and governance are obviously principles that are data platform independent and relevant to graph, relational, document... all DBs!