How this information was retrieved from the existing infrastructure to show all Tibco dependencies - was it done in automated manner or manually?Data retrieval is all done automatically as part of the connections that are established to the sources via the Stardog connection interface. Initial mappings can be automatically generated by may require some additional tuning to ensure it reflects the intended business ontology.
Source systems when critical may not allow connections for performance and security reasons. Doesn’t that affect the construction of an EKG? Can a EKG be used for calculating statistics efficiently? Where does the EKG position itself in the enterprise architecture landscape?As far as EA, we typically sit between the storage, governance and analytics layer, running on any VPC or On-prem infrastructure. Will need more information on your query regarding statistics...for ex: if you mean query statistics for cost optimization or statistics about the data (rolling averages, sums, etc).
KGraph sounds a lot like Halpin's ORM, representing objects(things) playing roles in relationships with other objects. Can I formally define or label "roles" in a KGraph? Can I Represent ternary and higher order relationships? Can I represent a relationship on a single object playing two roles, e.g., Employee is the boss of Employee (Reflexive relationship)?Yes you can define or label "roles" in a knowledge graph. The relationships are represented as description logic using OWL standards. No you cannot represent ternary and high order relationships, but you can represent reflexive relationships.
For the inference , did you use OWL or Rules ? Can you explain atomic triples? I'm not familiar with that term.Non-infered data representation is what I meant by atomic triples.
So if you are representing/virtualizing a relational database store, you break out relationships for each "attribute" related to the Entity.Yes if it makes sense in the context of the business ontology.
How is the EKG related to emerging paradigms such as data mesh or data fabric? Thanks for the presentation! Why call it ”property graphs (i.e Neo4j)”? Isn’t the most common name graph database? That is the name used by DB-Engines.com.Labeled Property Graphs or Property Graphs are distinct type of Graph Databases that are different from Semantic Graphs like Stardog, which are all based on RDF. Neo4J is one of many Property Graphs. I was trying to make a generic distinction between the two types.
Do you have a connector for Qlik Sense?
Is a triple db the same as a graph db in this sence?
Is de graph directed e.g. subject, verb, object? How does this relate to knowledge rules (e.g. if A then B is not if B then A)?
What is the best way to modele the concept of “time” in a knowledge graph. Entities and their relationships can change. What if I am interested in what can be inferred from the evolution/history of my entities?
Is a data virtualization solution required to enable StarDog, or is it just a separate integration approach? I.e., do you need DV for StarDog or is StarDog DV capable itself?
Data Virtualization is not required, although most of our clients find that easier to get started without having to move their source data. Stardog is DV capable as one of many unique differentiating features of our Enterprise Knowledge Graph Platform.
Why would you want to represent RDF data as a LPG? - ie a bit more understanding on the benefits/challenges of both formats. Vendors from both areas advocate their benefits - is this a phoney/holy war?
Yes, it's a holy war, that is now phoney. Given that edge properties are supported as a first class thing, the two types of graphs are isomorphic to one another, anything you can represent in one you can easily do in the other. One of the biggest benefits of semantic graphs - it also offer semantics on top, which is not supported via LPG
What are other training and learning option. Is there a paid training?
How does Stardog expose the column and row attributes of a legacy application which isn't covered by a connector?
Data from the legacy application would either need to be staged somewhere or materialized, that is ingested in Stardog. Our Solution Architects have worked with clients to help connect their legacy data structures and represent that within our Enterprise Knowledge Graph.
How does inferencing work when your underlying data changes. Do we need to re-do the virtualisation in order to infer new knowledge?
Can we persist newly inferred knowledge back to the database?
Could you explain, how did you infer overburdened resources? Is it based on some counts?
Stardog uses built-in reasoning capability to infer logical relationships along with rules. In this case the count of resources over a certain threshold was factored into what relationship constitued as over-burdened.
What are some of the ways latency is mitigated when using data virtualization in an enterprise KG?
Does that mean different domain different industry will have different knowledge graph?
Yes, typically, although, knowledge graphs can be subject area based like Sales, Marketing, Service and in that context could be shared across industries.
Do you have a connector for Qlik Sense?We can publish data in Stardog as a SQL End-point and most popular BI tools can communicate with Stardog in that manner. Assuming Qlik sense operates the same way, it should work.
Is a triple db the same as a graph db in this sence? Is de graph directed e.g. subject, verb, object? How does this relate to knowledge rules (e.g. if A then B is not if B then A)?It is directed. Knowledge rules in Stardog are based on SWRL (Semantic Web Rule Language).
What is the best way to modele the concept of “time” in a knowledge graph. Entities and their relationships can change. What if I am interested in what can be inferred from the evolution/history of my entities?It's dependent on how you'd like to analyze the data. You can use something like prov-o to capture the change history and/or you can use edge properties to annotate data and include timestamps, and then filter appropriately. Or you can take even more high level modeling approaches
Is a data virtualization solution required to enable StarDog, or is it just a separate integration approach? I.e., do you need DV for StarDog or is StarDog DV capable itself? Why would you want to represent RDF data as a LPG? - ie a bit more understanding on the benefits/challenges of both formats. Vendors from both areas advocate their benefits - is this a phoney/holy war? What are other training and learning option. Is there a paid training?We have Stardog Academy available for free. Hands-on training with tutorials can also be provided and customized for your needs - https://www.stardog.com/learn-stardog/
How does Stardog expose the column and row attributes of a legacy application which isn't covered by a connector? How does inferencing work when your underlying data changes. Do we need to re-do the virtualisation in order to infer new knowledge?This question is exactly why we do reasoning/inferencing at query time and we do not precompute anything. There is nothing to redo when data changes. All queries using reasoning will always use the most up to date data.
Can we persist newly inferred knowledge back to the database?Yes. It can be persisted back in the database, especially useful when employing statistics based inferencing using machine learning.
Could you explain, how did you infer overburdened resources? Is it based on some counts? What are some of the ways latency is mitigated when using data virtualization in an enterprise KG?Latency is mitigated through query plan optimizations, caching and clustering strategies.
Does that mean different domain different industry will have different knowledge graph?
