How to use Dgraph to design temporal knowledge graph?

Users Dgraph
1

The requirement is to design a knowledge graph in which some entity/predicate has temporal feature.

Consider an example like Linkedin.
Alice worked at company Microsoft since 2002-02-02, and this is her first job;
And Alice worked at company Amazon since 2012-02-02, and this is not her first job;

How to design with Dgraph to store all of Alice’s jobs and easy to query at which company did her work at specified time?

I think about using facets as below, but the old data will be lost.
Do you have any good idea? Thank you in advance.

{
set{
_:alice <dgraph.type> “test.Person” .
_:alice <test.name> “Alice” .
_:alice <test.gender> “female” .
_:alice <test.company> “Microsoft” (since=2002-02-02T02:02:02, first=true) .
}
}

{
set{
<0x2e794> <test.company> “Amazon” (since=2012-02-02T02:02:02, first=false) .
}
}

2

Hi @wanglj
In my opinion a more flexible structure is to have a type for companies and one specific type for job history. I’m using GraphQL type here and you can use dql equivalent:

type Person {
  PersonID: ID!
  name: String
  workHistory: [WorkHistory]
}

type Company {
  CompanyID: ID!
  name: String!
  workHistories: [WorkHistory]
}

type WorkHistory {
  WorkHistoryID: ID!
  company: Company!  @hasInverse(field: "workHistories")
  person: Person @hasInverse(field: "workHistory")
  date: Date
}

Now with this schema I do not need to store which work place is the first and I can get the result using a query.

What you can get easy with this structure(LinkedIn Example):
1- First place somebody worked at:

query {
  getPerson(PersonID: "0x2") {
     name
     workHistories(first: -1) {
        WorkHistoryID
        company {
          CompanyID
          name
        }
     }
  }
}

2- Workers of specific company:

query {
  getCompany(ComponayID: "0x1") {
    CompanyID
    workHistories {
        WorkHistoryID
        person {
            name
         }
     }
  }
}
1 Like
3

Thank you @pshaddel
I think it is a good idea, I will give it a try.