What does an Ontologist do?

If you’ve been following our blog posts, you’d now have an idea of what ontology is. But now you might be wondering, what does an Ontologist actually do? This blog post will aim to answer that by focusing on what we do here at SEEK. You will find that the role of an Ontologist varies from company to company, just like many other roles do! Now, let’s get into it.

Subject Matter Domain Knowledge

One important part of the job is understanding the domain that you are trying to represent or model. Each Ontology will be focused on a certain domain and in order to properly represent that area, you will need to be either already knowledgeable of, or spend time to become familiar with the domain. When moving into an ontology related role, subject matter expertise is an extremely valuable attribute.

An Ontologist can either gain subject matter knowledge through analysis or research, prior knowledge of the domain (for example, having previously worked in the field) or through engagement with experts in the domain. Learning about the domain isn’t something you just do once; it’s a continuous learning process. As an Ontologist, you are constantly learning about new things in your domain as new information emerges. As your Ontology grows, so does your knowledge of the domain.

Additionally, possessing domain knowledge is impactful because it helps to reflect how users think of the space. This domain knowledge is beneficial for providing context that supports decisions when it comes to Ontology modelling and curation. Having a team with different experiences and knowledge can be a key factor to a successful asset that allows for reuse across multiple markets or use cases. Diverse team knowledge helps to shape, deepen and create the Ontology so that it is representative of the domain you are working in.

Understanding User/Business/Product Needs

Now that you have a handle on the domain knowledge, you now need to gather an understanding of the user, business and/or product needs for your Ontology. This is where it’s important to engage with stakeholders and analyse what their needs are, so that you end up with an Ontology that meets the requirements of the end consumers. Of course, different stakeholders will have different needs, and this is where an Ontologist should be conscious of how their modelling decisions will enable flexibility and reusability across different use cases.

However, understanding and engaging with stakeholders is a two-way street. Stakeholders also need to understand what Ontology is capable of and what some of its constraints are, so that a mutual outcome is reached. Commonly, this means having people from a range of disciplines, such as Project Managers, Data Scientists, UX Designers or Subject Matter Experts, involved in on-going conversations so that a common understanding is reached. Having awareness of the business needs can also help the Ontology design to be reusable across use cases.

Ontology Design

The next big part of what an Ontologist does is the design of the Ontology, or what we sometimes call ‘Ontology modeling’. Ontology design is informed by stakeholder requirements and knowledge of the domain. When determining what the design might look like, an Ontologist will need to think about current and future use cases to ensure that the ontology design is reusable and flexible to meet the requirements of each use case. These decisions should support all use cases and ensure that efficiency is built into the Ontology design. Some of things an Ontologist will consider during this process is:

• What are the use cases and what are the requirements?
• What are the boundaries of the Ontology? This entails distilling down the domain into segments to work out what will and won’t be contained within the Ontology
• What are concepts boundaries? This is about working out what the concepts really mean
• What relationships are needed? This is where you start to think about how concepts are related
• What are the classes and entity types that are needed? So what types of things the concepts are
• What attributes are required? So, this is where you will consider what important metadata you need to capture about the concepts.

However, once the initial design is complete, this won’t be a set-and-forget. As new use cases or knowledge emerges, there will be a need to revisit the ontology design and see if it still fits its application. The revision of the ontology design is a constant in ontology practice.

Building the ontology

Perhaps the most obvious but important task an Ontologist will be concerned with is building the Ontology. There are many ways to build out an Ontology. Some approaches might include:

• Automating the build
• Human curation which is composed of research and use of predictive algorithms to support decisions
• Using machines with a human in the loop

But you don’t need to use one approach, sometimes it’s suitable to use a combination of them. The approach you take will be based on a range of factors, such as what technology is available to you; the scale and complexity of the data available; stakeholder needs; or whether it’s an initial build out or an expansion of an existing Ontology.

With each of these decisions, you will have trade-offs for scalability, efficiency and data quality. Thought also needs to be given to the maintainability and efficiency of growing the Ontology. Often, we find that one way of building out the ontology is suitable for a certain use case while others require different methods. So, as new applications for the Ontology develop, processes for growing the Ontology will change. Additionally, as technology changes and the discipline evolves, new methods for building an Ontology will develop and can be incorporated into your practice.

Maintaining the Ontology

As new information and use cases emerge, the Ontology will undoubtedly change (remember, it’s not a set-and-forget). This is an important part of what an Ontologist does — thinking about how to maintain the freshness of the Ontology over time. And also how the design can be refined to enable new use cases or improve efficiency and impact. Discovering new or missed information is an ongoing process. An Ontologist will think about how to discover this information and the types of feedback loops that are needed.

These feedback loops can help to uncover what is missing or any changes that need to be made to the Ontology. The type of feedback loops you implement will really depend upon the domain of the Ontology and what data resources or signals are available. This helps to identify new concepts or relationships that weren’t previously captured, and track changes in the domain.

At this point, an Ontologist would be thinking about developing tools or processes that help identify logical inconsistencies or where something within the Ontology is not in line with the design. It’s about making sure that the data within the Ontology is up-to-date and representative of the domain and is consistent with the Ontology model.

Developing Ontology metrics

Another important aspect of the job is developing metrics that measure the impact and value of the Ontology. Some of the reasons why an Ontologist might decide to build metrics include:

• Determining areas of improvement for the Ontology
• Prioritising where to focus our efforts
• Demonstrating the value that Ontology adds to the business
• Making evidence based and data backed decisions for the Ontology
• Helping to set benchmarks and markers of success
• Assisting with Stakeholder project collaboration.

As you can see there are many benefits to developing Ontology metrics and thankfully, sometimes the process can be straight forward — for example, counting how many concepts, relationships and labels make up the Ontology. However, this quantitative approach only tells part of the story and sometimes a different approach is needed to determine the quality, effectiveness and performance of the Ontology.

This is where it can become particularly challenging, as the application of the Ontology will dictate the complexity and availability of metrics. For example, Ontology can form part of a complex chain of services that feed into each other, which also have additional logic built on top. This complex system can make it challenging to determine the impact of the Ontology. It is at this point that an Ontologist may work with other stakeholders to figure out how and what metrics to develop for different use cases.

To conclude

There are many things that an Ontologist does to ensure they’re developing an Ontology that is fresh, scalable and fit for purpose. What an Ontologist does will change from organisation to organisation. However, hopefully you have a better understanding of some of the typical responsibilities and what role an Ontologist plays. If you’re feeling encouraged to learn more about Ontology, look out for our next blog on comparing different use cases for knowledge graphs, semantic graphs and ontologies.