The Knowledge Graph Conference Icon
The Knowledge Graph Conference
  • 🏠Home
  • 📅Events
  • 👤Members
  • 🔵Announcements
  • 🔵Ask
  • 🔵Ask The Ontologists
  • 🔵Events
  • 🔵Jobs
  • 🔵Promotions
  • 🔵Share
Powered by Tightknit
Share
Share

Understanding Knowledge Graphs: Definitions and Key Differences

Avatar of Riccardo De LuciaRiccardo De Lucia
·May 04, 2021 02:21 PM

Hi to everyone! As a newbie in this field, I have a really basic question about Knowledge Graphs. How would you describe what a Knowledge Graph actually is? What are its differences wrt a simple Graph? For instance, when we work with tools such as graph databases, we are  basically dealing with graphs. Then,  these databases can be used to model a knowledge graph, but their core is not inherently a knowledge graph. So, what does it take to a graph to be defined as a knowledge graph? To the best of my comprehension, a graph becomes a knowledge graph when its fundamental elements have semantic meaning. When a node represents an entity with a meaning, and when those nodes are related with a meaningful relation to humans, this is a knowledge graph. Even at this point, I struggle at making a concrete example of something that could not defined as knowledge graph, since almost every concept we, as humans, tend to connect has a meaning to us. Can someone help me clearing these concepts? Thank you!

👍3

7 comments

· Sorted by Oldest
  • Avatar of Ellie Y.
    Ellie Y.
    ·

    Matthias S. François S. Paco N. Phil T. Bojan Božić Aaron B.

  • Avatar of Bojan Božić
    Bojan Božić
    ·

    Hey Riccardo, I would say the simplest explanation would be: if all edges in your graph represent the same relation, it’s not a knowledge graph. So there needs to be inherent meaning that is represented in a graph. Your intuition was absolutely correct, as what we’re looking for in a knowledge graph is semantic meaning aka knowledge. Let’s say you have a graph representation of a network topology, this would not be a knowledge graph as every edge represents the same and there is no inherent knowledge that could be derived from it (only information). Hope that makes sense.

    👍2
  • Avatar of François S.
    François S.
    ·

    This all sounds good. The amount of semantics associated to information represented as a graph can vary significantly. Remember the term was initially coined by a marketing team at Google. A KG is not defined mathematically, although researchers in the field have tried to categorized them as precisely as possible (see https://arxiv.org/abs/2003.02320)

    👍2
  • Avatar of Phil T.
    Phil T.
    ·

    Most people agree that the distinction comes down to the semantic addition; because this facilitates understanding, which is generally considered requisite for knowledge and learning. I would go a little further and say that to be useful, the meaning needs to be shared and agreed (even if informally). If we have shared agreement on semantics then we can collaborate, make assertions and inferences (form new or virtual relationships etc), produce outcomes... On your last point, and to amplify Bojan Božić’s excellent answer, if your connected graph isn’t particularly useful, somewhat subjectively it probably shouldn’t be considered a knowledge graph.

    👍1
  • Avatar of Pascal H.
    Pascal H.
    ·

    It may be helpful to try understand some historic context, as there is no "crisp" agreed-upon definition. E.g. have a look at https://cacm.acm.org/magazines/2021/2/250085-a-review-of-the-semantic-web-field/fulltext, or Juan S.’s https://cacm.acm.org/magazines/2021/3/250711-knowledge-graphs/fulltext

    👍4
  • Avatar of Juan S.
    Juan S.
    ·

    KG is a modern manifestation of an ultimate vision in computer science: integrate knowledge and data at scale. It happens to be done using a graph data structure

    👍4
  • Avatar of Riccardo De Lucia
    Riccardo De Lucia
    ·

    Thank you all for your really useful and meaningful answers! 😄

💯
2
🙌1
🙌3
☝️2