Data model

Overview

Edge handles the complexity of your data by converting it into a knowledge graph that you can analyse and navigate quickly via simple no-code queries.

We use a knowledge graph data structure as it allows you to find hidden patterns with business value that would take orders of magnitude more time and effort to unearth if you were to use traditional analytics methods on each individual dataset at a time.

Why a Knowledge Graph?

• Knowledge Graphs provide single source of truth for all data, virtually eliminating database refactoring.

• Knowledge Graphs solve data sparsity by storing necessary information without needing extra space for all the features that aren’t necessary.

• Knowledge Graphs solve data heterogeneity as they allow diverse datasets (videos, images, sensor feeds, structured and unstructured text, etc.) to be fused together. This integration enables the extraction of insights that transcend individual data silos, treating them as a unified entity - one cohesive knowledge graph.

Data Types

In this tool, data is encoded as Nodes and Edges, and organised through Views.

Nodes

Nodes can represent a variety of different ideas and semantic meanings. For example, some nodes represent individual data points, whilst others represent the features that those data points contain. Let's look at some examples of data types which nodes can represent, and how we can interact with them:

• TEXT Some nodes represent chunks of text, the content of which is stored in their node labels - you can read the full label as well as toggle label visibility by opening the Node(s) -> Properties window.

• MEDIA You can watch, or view, any media which is stored in the node by highlighting a node then using the Watch Media button within the Node(s) menu. If the button is greyed-out, the node does not contain media.

• URL/ FILE Some nodes contain links to web pages or files, stored in their "URL" node property (you can view all node properties in the Properties window, accessible through the Node(s) menu). You can open the URL attached to a node by highlightng the node and then selecting Open URL/ File in the Node(s) menu.

Taxonomies

To organise, analyse, and navigate through the knowledge graph, we use a taxonomy of nodes which represent features in the dataset. Such nodes are also known as "header nodes" or "annotations".

When an annotation connects via an edge to another data point, it tells us that the data point contains this feature. For example, in the case of an Audit database, there will likely be an annotation with label "CANCELLED" that connects to all audits which got cancelled.

Edges

In the example we just gave above, it was the edge pointing from the "CANCELLED" node to the audit node which tells us that this audit was cancelled—edges record relationships between nodes.

In Edge, the edges are directed (they point from one node to another) and have a weight which stores numerical values. For example, in the case of a database of scenarios used for testing an autonomous vehicle there may be a header node with label "number of cars" - this will connect to all the scenario nodes via edges whose weight is an integer that tells us how many vehicles were involved in each scenario.

Views

To interact with the data inside our knowledge graph, we use views. You can think of each view as a 2D slices of our knowledge graph; a way of peering down onto your data from different angles, looking at only subsets of the knowledge graph database at a time. In Edge, you can open multiple views at once which contain the same data points, but aligned along different axes. Looking at the same data from different angles allows us to reveal and understand hidden patterns within our dataset.