This makes the remarkable claim that the ontology document is correct no matter what the actual data says! And I suspect legal would have something to say about the ontology defining what a contract is.
It badly needs grounding in some real examples. It reads like pure navel gazing academia, or expensive, gold plated consultancy.
I have never encountered a company that talks about their ontology.
You need the warehouse.
You can benefit from a semantic layer. Optional, but loads of examples in the wild.
Definitions in a semantic layer are absolutely an ontology of sorts. You should avoid working with people who let conceptual purity get in the way of practical results. Concepts and metaphors can guide your work and help with communication; pedantry drags it to a halt.
Treating your ontology as not just a useful concept but an actual product you spend time writing and maintaining? Outside of it being a word you could reasonably apply to tools such as data catalogs, my anecdata is that I have never heard of it in practice.
Ontologies have become much more mainstream over the past year or so. Case in point: Microsoft's new Fabric IQ product, https://learn.microsoft.com/en-us/fabric/iq/overview. I don't know how good this product is, but there's a LOT of interest in the space.
Although I agree with the separation being argued here I don’t find the practical implications that helpful especially in an age where LLMs will be writing a proportionally increasing amount of queries. In the same way a semantic layer can be inconsistent with itself, a semantic layer can be inconsistent with a written ontology that doesn’t technically tie to your actual system. What are the actual best ways that ontologies have been implemented? Ive heard of how it works at Palantir but are they the exception?
This makes the remarkable claim that the ontology document is correct no matter what the actual data says! And I suspect legal would have something to say about the ontology defining what a contract is.
It badly needs grounding in some real examples. It reads like pure navel gazing academia, or expensive, gold plated consultancy.
I have never encountered a company that talks about their ontology.
You need the warehouse.
You can benefit from a semantic layer. Optional, but loads of examples in the wild.
Definitions in a semantic layer are absolutely an ontology of sorts. You should avoid working with people who let conceptual purity get in the way of practical results. Concepts and metaphors can guide your work and help with communication; pedantry drags it to a halt.
Treating your ontology as not just a useful concept but an actual product you spend time writing and maintaining? Outside of it being a word you could reasonably apply to tools such as data catalogs, my anecdata is that I have never heard of it in practice.
Ontologies have become much more mainstream over the past year or so. Case in point: Microsoft's new Fabric IQ product, https://learn.microsoft.com/en-us/fabric/iq/overview. I don't know how good this product is, but there's a LOT of interest in the space.
Although I agree with the separation being argued here I don’t find the practical implications that helpful especially in an age where LLMs will be writing a proportionally increasing amount of queries. In the same way a semantic layer can be inconsistent with itself, a semantic layer can be inconsistent with a written ontology that doesn’t technically tie to your actual system. What are the actual best ways that ontologies have been implemented? Ive heard of how it works at Palantir but are they the exception?
I've had some difficulties understanding the term ontology when I first heard about it, but now it starts to make sense.
Btw, tropes.fyi/vetter says the article is sus. Perhaps LLM written. But I see no disclaimer.
The concept of an ontology is really simple: thing - relationship - thing. Example: peanut butter - goes well with - jelly.
This is in contract to a taxonomy, which is hierarchical:
(upper-level category): Sandwiches
(sub-type): peanut butter & jelly
(sub-type): grilled cheese
(sub-type): BLT
(upper-level category): Soups
(sub-type): tomato
(sub-type): chicken noodle
Taxonomies are kinda boring because they're just hierarchical lists.
Ontologies are cool; the relationships are usually the important thing.
Noted, thank you.