There are three different types of relationships in taxonomies:
Equivalent (Synonyms: “International Business Machines = IBM”)
Hierarchical (Parent/Child : “Computer Manufacturers => IBM”)
Associative (Concept/Concept: “Software Group – Software”)
Heather Hedden’s presentation on taxonomy powered discovery for a recent Boston KM Forum contained an interesting set of examples for how to organize the last type of conceptually related term sets.
Process and agent: Programming – Programmers
Process and instrument: Skiing – Skis
Process and counter-agent: Infections – Antibiotics
Action and property: Environmental cleanup – Pollution
Action and target: Auto repair – Automobiles
Cause and effect: Hurricanes – Flooding
Object and property: Plastics – Elasticity
Raw material and product: Timber – Wood products
Discipline and practitioner: Physics – Physicists
Discipline and object: Literature – Books
Associative relationships are the mechanism behind a search interface that presents related content to a user. “You might also be interested in…” By building out these kinds of conceptual relationships, it’s possible to anticipate what a user wants (if we know who they are and something about what they are doing).
This is also one way to turn a taxonomy into an ontology. An ontology is simply a set of terms and term relationships that represents a body of knowledge. For the life sciences industry, this might include biochemical pathways, genes, mechanism of action, generic names of drugs, chemical compounds and so on. For the insurance industry it might mean products, risks, regions, policies, claims processes, etc.
Developing the associative relationships between terms allows users to locate the information needed for a task. If I am a researcher, I might want to find current antibiotics in development for an infectious agent. If a claims processor I might want to know contributing factors to an accident.
These types of relationships require human judgement to be applied to your content, your users and your processes. What is important? What related information could help someone make a decision?
The advantage of mapping these semantically is that content in multiple systems can then be related and the relationships can be changed quickly without having to reorganize the actual content at the repository level.