I'm sure this won't win me many friends in this group, but I wrote an essay about the risks of knowledge graph hype 馃槄 and offer up an alternative view of how we can grow the impact of KG technologies: https://blog.diffbot.com/from-knowledge-graphs-to-knowledge-workflows/
thanks Mike, I like how you frame this around automating knowledge workflows. we are also planning an industry survey and this article is helpful context for that.
Agree with this:
. A much better model is to have federated, distributed workflows that each team owns, keeping small amounts of internal state and tapping into automatically updated streams of external data to get the latest information at the point in time its used.
What are your thoughts on Data Mesh: https://martinfowler.com/articles/data-monolith-to-mesh.html I think combining KG with the architectural ideas of data mesh is the way to go (and I've already started to see this pan out)
Cool, I haven't heard the term Data Mesh being used and will read this. I guess my main point there is that different teams in a company are already federated and they don't need a company-wide data architecture, which sets us up for the trough of disillusionment. If we study what information actually needs to be shared across teams, KG tech can be used here to streamline that by making sure those high-throughput information flows are structured data. we can also have impact much faster, as measured by actual productivity.
Its stuff like this whitepaper that prompted me to write the essay. I think its bordering on malinformation that doesn't help anyone over the long-term: https://www2.deloitte.com/content/dam/Deloitte/de/Documents/operations/knowledge-graphs-pov.pdf
Our next catalog and cocktails Podcast will be demystifying data fabric and data mesh. Should be fun. We are recording it live today, feel free to join. Or listen to it tomorrow https://data.world/resources/webinars/catalog-and-cocktails/
From an honest, no-bs perspective, the data mesh has a lot of promise. It鈥檚 actually less about new technology and more about organizing the process, people and culture
Agree strongly with this and really like the blog. The combination of positive organisational shift and better data ownership with an enabling knowledge graph is a really powerful set of ideas. Happy to see it coming up here.
Great definition - will spread and promote! 'Knowledge Graph technologies allow the introduction of automation into information worker workflows, helping them save time on mundane information processing tasks.'
Thanks for sharing Mike. I share the views you expressed in this post. Thanks for the links on the data mesh Juan, I will look at this closely.
Mike Tung Listen to our podcast episode on Data Mesh: https://podcasts.apple.com/us/podcast/is-your-data-fabric-a-mesh/id1524652737?i=1000506100424 There are a lot of dots connected with your post.
hey Juan S.! i listened to this on the way into work. Sounds like a Data Mesh is a model where each domain-specific team is responsible for serving their own data (distributed serving), but there is a centralized agreement on vocabulary/standards.? Kind of like the web? I think the tough pill to swallow is this centralized definitions/standards committee.
Mike Tung You got it! Agreed that the centralized definition standards committee can be a hard one, but the way I think about it is that you want to centralize the core model of your organization. For example, no one should redefine what is an email, first name, address, etc. Let the domains extend the core model. Sooner than later, people will complain that different domains are re-defining the same concepts and that will get reported to the centralized committee and they will take action. So part of the paradigm shift is to be comfortable with friction. It's never going to be perfect, but as time goes, the core model will expand and get better. Also note that if a domain doesn't use the core model, the data consumers will complain. Finally, schema.org is an example that you can have a centralized standards committee that grows little by little and gets adoption.
Ideally, yes. In reality, look at any company's CRM. Even within a team of people that has the same goal (sales), they struggle to define common definitions, and more importantly, enforcing them. So I think we have to be realistic, from the perspective of the one dealing with the data, they have to consider it a net productivity gain, otherwise it's just bureaucracy.
Yes i think schema.org is a good example of adopting some structured data incrementally into the workflow (i.e. the Wordpress/CMS workflow). It's not an internet-wide data interchange format though--rather it works because its controlled effectively by the search engines and people conform to it because they want to get ranked in the search results.
schema.org is a way for you to annotate your content for Google
