21st June: Introducing Ontologies

Summary of the First Round Table. 21.06.2022:

The first session of the series “Round Table of Research Data Management” welcomed numerous participants from various disciplines and was aimed at starting a discussion on fluid ontologies as a key aspect and general outcome of the Digital Research Environment (DRE) of the Cluster. It is clear that fluid ontologies require more than a purely technical solution. For this reason, the Round Table would like to exchange ideas about fluid ontologies. Therefore, the Round Table would like to encourage the exchange of ideas and collect the various approaches to what fluid ontologies can be. The goal is to design a system within DRE that is open to all researchers and assists researchers. It should also foster new ideas. To do so, we have to find and develop a common understanding of what fluid ontologies can be. This includes a common vocabulary for these concepts. The first thing to do is therefore to define what ontologies are.

From a Library perspective, as demonstrated by Wynand van Der Walt, there are key aspects that need to be considered when defining what ontologies are. Firstly, how do we organise knowledge, and what tools do we use: thesaurus, taxonomies or ontologies? These tools are not interchangeable but instead represent various approaches toward accumulated knowledge and how it is ordered and represented. Secondly, the definition of ontologies is shaped by distinctive rich relationships between language terms, mostly natural language. We therefore need to know how it is being translated into a digital multifaceted environment. When we try to organise knowledge, we additionally have to address decolonisation, that is, the marginalisation through particular organisation processes. We need to know for whom we are organising knowledge, ensure representativeness of indigenous knowledge systems and acknowledge the plurality of knowledge systems or alternative structures. The abstract concept of ‘Ubuntu”, a Zulu concept, shows for example how complex an ontology would have to be to reflect the complexity of this particular term. ‘Ubuntu’ incorporates various understandings that do not have direct translatable equivalents in other languages, including English. The ontologies thus need be interpretable by humans and machines alike and provide the relevant meanings that such concepts carry.

From a computer science perspective, an ontology is a formal, explicit specification of a shared conceptualisation (Gruber 1995). Once this is defined, it is modelled, that is, representing information in a particular way for a very specific purpose. The humans that have made the decisions to include or exclude particular representations of the world determine the integrated information. Thus, ontologies can never be complete or represent the world comprehensively but are part of every knowledge-based system such as databases or artificial intelligence systems. It is clear that an ‘ontology’ needs to be a shared conceptualisation, which is why a consensus is needed to come to a mutual understanding of the concept. All concepts also need to be explicit. By that we mean that any concepts included in a model need to be clearly defined to be differentiated and be machine-readable. This is achieved by a methodology from which emerges terminological knowledge (design-time), defined by classes, concepts and relations, assertional knowledge, which contains knowledge at some point in time (run-time), and inferred knowledge, that contains combinations of classes, relations and constraints.

Discussions between participants allowed for the various understandings of the term ontologies to emerge. This included discussions on temporality which maintains that the term does not represent the changeability of things or assumes that ontologies, as we have discussed, are taxonomies. Taxonomies are however hierarchical, whereas ontologies can describe all associations that are not hierarchical but associative. Some understand that fluid ontologies suggest processes, such as interactions and relationships. Others see the opportunity in ontologies to connect to others and expand the connection of various associations and relations between singular ontologies.

Our understanding of ontologies should not be restrictive but rather be a multifaceted environment. Taxonomies will be part of the knowledge graph (more about knowledge graphs in the second Round Table meeting) that we are trying to create. The terminology we are using may be incorrect but we all agree that what we want to ensure is that the knowledge that we use is representative of all the disciplines within the cluster and can establish and confirm relationships between the emerging knowledge structures. The use of the word ontology is misleading and means in this instance a structured high-level taxonomy. We should be aware of the limitations of words and must thus choose and define them wisely.