The fundamental challenge in the discipline of knowledge management is that knowledge captured and represented outside of the human brain is never quite as accurate, reliable, or easy to connect as knowledge inside the human brain. Short of bequeathing actual brains in knowledge transfer sessions à la Spock’s Brain or a Vulcan mind meld, the best way we have to transfer knowledge is through the written word.
Writing down knowledge in the form of stone tablets, papyrus, paper, and pixels is what historical knowledge transfer has always been about. Capturing knowledge and then making it available to others via written words or video is really the only practical way to pass on knowledge at scale. It is overwhelming to consider how much knowledge has been lost over time due to knowledge not being recorded or destroyed by time. Who knows what Nikola Tesla invention we may be using today if he had documented all of his ideas and creations.
The problem for the current age is not how much knowledge has been lost, but how much knowledge has been created in the electronic age and and our ability to handle and make sense of all the information created. As we abstract information into consumable and processable bits, we lose context, meaning, and, probably, knowledge. In the computer age, we attempt to model information in ways which are understandable by users. Part of this modeling is using machine learning, artificial intelligence, and knowledge graphs.
Despite the ups and downs of hype cycles around new technologies and trends, the recent popularity of knowledge graphs has ramifications for information retrieval and question answering. The use of knowledge graphs has grown for several reasons. First, the amount of data accessible and connectable has grown as has the ability to identify it and place it in context. Second, the nature of social networks necessitates the use of graph structures rather than one way or reciprocal lines between individuals and facts about those individuals. Finally, the demands for getting insights and answers out of existing information has expanded. People expect more out of the information to which they have access.
The real appeal of knowledge graphs to knowledge management professionals, however, is not the trending popularity but the real-world implications of modeling information more closely to the way knowledge is structured. Because unstructured information–actually, all information has some structure–is the most answer-rich yet the most difficult to parse and extract answers from, the idea that this type of knowledge can be connected and presented in a way closer to the way in which we think and answer questions is of great interest to knowledge management.
Inherent in the rise of knowledge graphs is the rise of ontologies.
The general definition of “ontology” is “a branch of metaphysics concerned with the nature and relations of being”. When attempts to model domains of knowledge made the leap into the electronic world, the term stuck and has come to define structures both hierarchical and relational. In the world of knowledge modeling and organization, an ontology represents and models a domain bound by rules which determine what can be logically put together. True ontologies are modeled in triples, which include a subject, predicate, and object. Essentially, this mimics the structure of language and thought: subject, verb, object.
Put simply, ontologies include classes of things which can logically be grouped together. For example, people, organizations, document types, skills, etc. can all be grouped into individual classes because they share common attributes. It is then possible to connect specific concept instances (subjects and objects) within and between these classes using relationships (predicates) defining the nature of the connection between those concepts.
Within a class, you may have simple relationships like broader/ narrower to express a parent – child relationship. Between classes, you can create specific relationships defining the specific nature of the connection between concept items. Concepts can be connected with one-way or two-way relationships and they are not limited to the number of relationships between other concepts they can have.
A simple example:
Class:People Relationship Class:Books
Ian Fleming isAuthorOf Casino Royale
(Subject) (Predicate) (Object)
Class:Books Relationship Class:People
Casino Royale hasAuthor Ian Fleming
(Subject) (Predicate) (Object)
Ontologies Support Knowledge Graphs
What does this have to do with knowledge management? Ontologies model the domain, and, effectively, model the way we think about a domain. Once these ontologies are in place and provide the structure and specific instances of named items (as in any controlled vocabulary), they are then connected to actual “things” through tagging, such as documents, videos, databases, etc. are then queryable and actionable. Once they are actionable, they are knowledge graphs. These knowledge graphs support the capture and retrieval of all types of knowledge, including the information directly created by participants but also the metadata information associated with them and their work.
Knowledge graphs are a natural fit for knowledge management activities because they model domains (and cross-domains) of knowledge to retain more context and meaning even as information is parsed and abstracted for digital representation. Information is modeled in a way that is more intuitive and useful to users. Knowledge graphs do not replace traditional knowledge management activities; on the contrary, they are just one representation of the knowledge that is captured and assist users in accessing that knowledge.