You’ve probably used many solutions that have some kind of graph visualization built in, showing how things (entities) are connected (links). Entities and links may both have properties, which form the resulting network.
For basic visualization of networks these kinds of solutions help the user see specific elements and shapes of the network. However as soon as you start adding different data types, sources and properties these solutions often fall short and start to hinder the user at best, leading them to false assumptions at worst.
When looking to bring different data types together, or looking to start more complex analytics there are several things to look for in any solution:
Having the flexibility to work with any data, or any data type quickly and easily is paramount when carrying out analysis and investigations in a rapidly moving environment. Whatever solution you use should allow you to create or import unknown types and properties to quickly carry out visualisation and triage of the data and the network.
On a similar theme, once you’re happy with your data it’s important that you can align it to a known data schema, aka an ontology. By aligning your data to a known set of types, you can then run graph analytics easily and more accurately across the known types in your data set. Ontologies also allow for advanced features such as inheritance and semantics.
Very often the key requirement for bringing your data into a linked data analysis tool like i2 is to visualize a network and tell a story to your stakeholders. Not all visualization tools are created equally and it’s important your tool allows you to control the key elements and the design of your chart. It’s often said good design is CRAP – and allows you to account for Contrast and color, Repetition, Alignment and Proximity. Being able to control and customize the visualization of elements on your chart is key to presenting an accurate and compelling piece of evidence.
Knowledge graphs can be used both for visualization as well as analysis, and the type of styling and views you will use will typically vary on your application and requirement. Having a tool that allows you to dynamically change views or entity / link representations in an instant can be key. From showing individual to summarized linkages, or adding (or removing) key elements from entities on the chart or changing link styling, dynamic chart layouts allow users to quickly and easily make progress with their analysis.
As knowledge graphs get bigger, techniques to summarize and group entities and links become imperative to ensure that graphs are still understandable, and don’t just resemble a large death star / ball of string / insert your catchy metaphor here.
Entity grouping, entity aggregation and link summarisation are all techniques that can allow for similar items of interest to be grouped effectively to reduce duplication on the graph. Entity merging and / or entity union, where duplicates are either merged or hidden can also help, particularly in busy graphs where a particular entity (e.g. a key address) appears many times.
When using these techniques, it is important to consider the analytical implications these may have and whether the tool you are using will take into account the underlying structure of the graph when looking at analytics such as find shortest path or social network analysis.
However good the tool you use is, at some point in working with your chart you are going to need a level of customization, perhaps for a new entity type or to call out a particular results set on your chart.
Being able to easily add new types, new icons, new rules and the like and share them amongst your organization is key to enabling both productivity and consistency in your analysis.
Semantics are an often overlooked but extremely powerful feature of knowledge graphs, particularly if you need to pivot between multiple data sources where the data types are similar but different in their ontologies.
By enabling semantic types, you can categorize specific entity types, e.g. a car is also known by its semantic type e.g. a vehicle. Semantics in essence allow you to generalize a type, which then allows you to select, search, pivot, add data and run analytics effectively across entities and links from different sources effectively, whilst maintaining their source properties and typing.
Often referred to as ‘data fusion’ is the ability to quickly and easily de-duplicate entities and links on your chart. This is a key requirement when carrying out any kind of analysis in your graph. Duplicate entities introduce errors into graph analytics and can render them difficult to interpret and unreliable.
A key, must-have requirement for any analyst should be the ability to track and understand the provenance, i.e. where it originated from. This should be evident in the entities, links and properties on your graph.
Being able to keep a reference or a link to the source of your information is key, and as further information is received being able to hold multiple source references on the entity, link or even property becomes paramount for when you need to explain how you have come to your conclusion.
Tools like i2 also allow you to follow a grading system on your intelligence to mark the confidence in your source, intelligence and even handling, a methodology critical for those working in Intelligence Analysis in Law Enforcement or National Security.
often the properties or attribute you hold on your entities and links will be one of the standard types such as text, number, boolean, etc.
However with analytical tools such as i2 Analyst’s Notebook, being able to hold special types that allow for either geo-spatial or temporal fields opens up a whole new level of analysis capabilities with your data.
It goes without saying that holding location co-ordinates, such as latitude and longitude allow you to place an entity or group of entities to look at the “where” on an event. Temporal properties such as date, time or datetime let you look at the sequence of events.
Combining these two properties allow you to look at the full sequence of events as well as their connections and networks, and can feed into special analytics such as co-traveller analysis; where you may look for two entities who come together in the same location for a specific period of time.
The above considerations differentiate between a tool that can visualise data and an application that is providing visual analytics. Understanding the difference between data visualization tools and visual analytics platforms is crucial to fully benefit from the advantages of a visual analytics application. These solutions offer a range of visualization options, enabling users to easily identify patterns and signatures within the data. A visual analytics application also provides a set of powerful visualization tools that allow users to: explore large datasets, identify trends and patterns and extract insights from complex data.
By providing an interactive environment for data exploration, and rich analytical capabilities, visual analytics applications empower users to make informed decisions based on data-driven insights.
This article was co-authored with Director of Research & Development, Adam Etches.
Adam Etches is our Director of Technology at i2 Group. Adam is responsible for our research team and technical vision. With a strong background in analytics, Adam has developed a unique understanding of the application of technology to the Intelligence Industry.