Knowledge Engineering- Types of Knowledge Graphs

Check — Knowledge Graph Series [https://ogre51.medium.com/knowledge-graph-series-01aa961c0d21]

There are four types of Knowledge Graphs (KGs) we can create for our Knowledge Base (KB) based on what kind of data it has, and what schema we think suits it best.

So, we have to make a choice. And it depends on the use-case (how the KG will be used) and the general setup.

Two points need to be considered:

1. Is the data dynamic in the sense that is incurs regular/continuous updates; or
2. Is the schema dynamic (… developers choice based on the use case)

One can easily note that ‘fixed data and schema’ will be easier to build and maintain, whereas ‘dynamic data and schema’ will require tooling and workflows to keep graphs in sync with a growing corpus of data.

But let’s explore and be clear in our view. In general, in the case of:

• Dynamic Data and Dynamic Schema, we have to build many complex graphs that constantly grow.
• Fixed Data and Dynamic Schema, we have to iteratively experiment with schemas to build better representations of data, of a fixed KB.
• Dynamic Data and Fixed Schema, we have to stream data points in to the fixed schemas to collect opinionated structured information.
• Fixed Data and Fixed Schema, we need semantically accurate look ups, served by document hierarchies and chunk linking solutions.

Now, we may leave every argument aside, and just say that the first option is what we should go for. Dynamic data and dynamic schema, can sort of incorporate every feature offered by other Knowledge Engineering methodologies. But that may not be the ‘automatic’ choice to go for.

There are two primary reasons:

One, there are cases when the data has irrelevant information that we do not want represented in the KG. Having those irrelevant data points will make our retriever (and the process of graph extraction) perform poorly.

The idea is to represent data that’s relevant to the use case…