Ontology - a conceptual semantic model
The term 'ontology' (which can be off-putting) is understood as 'the study of being' or the 'logic of existence'. In this context, it applies to content - why does this content exist? The ontology acts as the organising mind: a generalisable framework for understanding entities, their context & relationships, using semantic tags. It is an active information architecture - not a taxonomy - and one which provides meaningful domain.
In defining entities and concepts, the ontology captures how meaning is inferred from relationships within a domain. This is a key differentiator from a taxonomy which doesn't seek relationships or context, just the next ancestor or descendant class. Since ontologies do not focus on classes, they are free to flex according to contextual, regulatory, organisational or use case changes. They can evolve where a taxonomy cannot.
Semantic tags are metadata, words or terms that typically describe web content on websites to search engines. In the context of knowledge graphing, semantic tagging is the process of associating an element from an ontology with a piece of content - even just a cell of data. Semantic tagging provides a descriptor, linked to a unique URI location, which enables faster retrieval subsequently. Semantic tagging is what the ontology uses to class the content (and similarly tagged content) within the domain, as a body of knowledge about a certain topic. Note the need to disambiguate between semantic tagging on websites, and semantic tagging in the context of knowledge graphing: same phrase, different context, different meaning!
You have choices about how you design and build your model:
- You can decide whether to bespoke your ontology or adapt one that is already and publicly available. Both approaches have advantages and pitfalls. Third party libraries can contain a lot of material that just doesn't apply to your organisation, or, they may not be practicable to update. If your needs are niche or specific, it may be better to have one bespoked for you. If you believe there is potential in exploring content on the web, for fresh discoveries, then a standard library may be a good idea. If not, then it may be redundant. Consider also whether you really want to use someone else's view of information classification. You may not. These third party libraries are also very large datasets in their own right.
- You have the option of reverse engineering your conceptual model by extracting likely metadata from your existing content or metadata management tool. Or if you feel that is no longer appropriate, then design afresh using the opportunity to build organisational alignment. If you choose the former option, you will need services to extract the concepts but you can steal a march by applying the semantic tags and graph references at the same time. The benefit is this will already start to capture the meaning of the relationship between structured data & unstructured content (text), generating meaningful connections between the unstructured text and the structured data. Using services in this way is also a good idea if you want to automate the process of applying metadata to what may be a considerable digital estate.