Ontology? What’s that?
Written by Hannah Nesbitt, Kate Mayhew and Simon Jaskiewicz.
So, you’ve started a new job and you hear the word ontology being mentioned but are not sure what it is. Or maybe you’re at a family event and your cousin’s new partner tells you they’re an Ontologist and the top results on Google haven’t helped with what to ask next.
Sometimes it can be hard to work out from the Wikipedia pages if an Ontologist is a philosopher, a physicist or something completely different. Don’t worry! We’ve got your back and are here to help you work out what on earth ontology is.
At this point you might be wondering why listen to us. We haven’t posted before, and we all know trust on the internet is lower than it’s ever been, so it’s probably a good time to introduce ourselves. We’re a collection of team members in SEEK’s Ontology team, a part of our AI and platform services department. We’ve been an integral part of the SEEK family for about 8 years now and although we haven’t written any blog posts in the past, we’re looking to change that.
This post is a starting point but keep an eye out because we plan to follow up with some more in-depth content. For now, let’s start with the basics and help you tackle that awkward family event.
It all started with philosophy, and now it’s applied to the world of tech
Ontology in the philosophical sense was first conceptualised in either 5th or 6th Century Greece by a Greek philosopher called Parmenides. It is a branch of metaphysics that tries to answer the questions of the existence and nature of being. In basic terms, it involves the questioning of what things exist, how they relate to each other and how to classify or group together such things according to similarities and differences.
Since then, theoretical ontological philosophy has proven invaluable when practically applied to help structure data in computer science. Ontological thinking provides a way to describe real world concepts, their properties and how they relate to other things in a way that’s interpretable by machines.
That idea might be a bit hard to relate to, so a real life example is the knowledge panels you get when you Google search. A search for a saxophone will give you different types of information in the panel than a search for a concert hall, because the kinds of things you might want to know about an instrument are different to what you might want to know about a venue. This idea can be expanded to solve all kinds of technical problems through ontologies.
A ‘thing’ is a concept
To be able to provide descriptions of things in the real world to machines, we create what we call a concept, or sometimes a node. A concept is a representation of a thing that has an identifier and captures all the knowledge we have about that thing against that one identifier. Still with us? So, for example, if we were creating an ontology representing musicians, Beyonce would have a concept that captures information about her, including that her nickname is Queen B, and that she has an alter ego of Sasha Fierce. By doing this we can treat all three names as the one musician.
Other terms you’re likely to hear us use include:
- Attribute — this is a piece of information about the concept. An example would be that “vox” is slang for “vocals”. Attributes can be text, numbers, or even a true or false flag.
- Relationship, sometimes called an edge — the way we connect two things together. These can either be weighted to tell us how close they are, or a named relationship that tells us how they are connected. We might relate “recorder” to “Bach”.
- Class — the type of thing a concept is. Using this we can say that a “trombone” is an instrument, and that “Chumbawamba” is a band.
An ontology can look like that window scene in the movie ‘A Beautiful Mind’ but it doesn’t have to be that overwhelming
When put together it can be a bit intimidating and it’s easy to pop it in the too hard (you lost me at ontology) basket. In reality though there are a lot of ways that we can simplify ontology assets for people to use. In practice, what this can look like is providing a slice of the ontology as various file formats, creating a table format or building tools that allow people to call the assets with just the information they have and get back the specific information they need.
You’re probably using ontologies without knowing it
You might be asking yourself “What does this thing look like?” and chances are, you encounter ontologies every day without realising it. Google, Amazon, Facebook, Netflix and Uber Eats all use Ontologies, and sometimes you are even helping to build them! Think about when you like a song on Spotify and see suggested similar artists.
It can be hard to show the ontology itself because it’s a data asset that sits behind systems to help them work but there are some examples where you can.
An everyday example that will likely be familiar is Wikipedia — their pages can be thought about as concepts, with attributes and links (relationships) to other pages (concepts). As a user, the structure of all the links allows you to look at everything that is connected, and if you like to get really nerdy about it (like we do) you can also use the data to create insights about what is in Wikipedia, such as musicians who can be found in each genre.
Ontology creates structures that enable you to do more with the data you have
Now this is not to say that ontology is the only solution when it comes to solving data problems, it all comes down to the use case and they’re often used in tandem with other technologies.
Ontologies are used to provide the extra bit of contextual data that is required to make a service or machine process go from good to great. Where AI services are the norm, ontologies offer a source of high quality, standardised, and reusable data, that can be interpreted by both humans and machines and used for a variety of purposes.
We’re hopeful that reading this has given you a clearer understanding of what ontology is and how it can help solve issues around data.
Keep an eye out for our upcoming blog posts (below) which will delve into how ontologies fit into AI, and the differences between Ontologies, Knowledge Graphs and Semantic Graphs (say what!?).
