As 2019 comes to an end we look back at what, if anything, has changed in terms of the trends in semantic technology and data governance as well as our thoughts on what to expect in 2020. Here are five top takeaways:
- Knowledge graphs are becoming ever more important.
This trend has caused providers of knowledge graph technologies (based on the semantic web standards stack) and solutions to have to distinguish knowledge graphs from other types of graphs, such as property graphs. There are many succinct comparisons and a common list of unique capabilities that knowledge graphs provide versus certain limitations of how data can be organized using some non-standards-based types of graph structures. The most important differentiator is that knowledge graphs contain not only facts (data) but also the meaning of the facts. The meaning is expressed in data schema (models) and in rich rules. Schema and rules representing the meaning are referred to as an ontology. Availability of the ontology as part of a graph is a definitional quality for knowledge graphs. Ontologies are business models in contrast to, for example, data stored in relational databases, which does not provide or function as a business model.
Knowledge graphs represent the meaning of the data in a standards-based way, and make it available at “run time”. They can be consulted about what they know. New knowledge can be dynamically added at any time, integrated with what is already there in a meaningful way, and immediately available to users.
- Knowledge Graph (semantic technology) approaches provide a powerful basis for Machine Learning and AI.
We expect this trend to continue and diversify in the range of applications. As per the above points, the interaction between AI/ML processes and knowledge graphs can be synergistic (a virtuous cycle) because new information can be added at any time and merged into the available “knowledge” (because it is integrated with the meaning of the information). And, AI/ML can now be applied at a high-level, to business concerns themselves vs. more at the data level – because knowledge graphs provide a business model of the organization. The new aggregate of knowledge can, in turn, result in further rounds of productive additions from AI/ML processing.
In the data governance (DG) area, there is a strong interest in automating or semi-automating the meaningful integration of collected metadata. AI/ML can be applied to some metadata to assist in the integration. Knowledge graphs also enable inferencing and rules-based reasoning which is very helpful in the (semi-)automation of some DG processes and generation of new knowledge through, for instance, data relationship discovery and exploration. Combining this with Machine Learning processes can leverage synergies of both technologies and deliver a powerful solution.
- SHACL, the newest standard in the semantic stack, is being widely adopted
SHACL is a data modeling language for describing and validating knowledge graphs, We can now confirm that adoption of SHACL has been widespread. Today, just two years after the standard was released, nearly all of the RDF graph database vendors support SHACL. It also has been quickly adopted in business solutions that are using knowledge graphs.
- Semantic technologies are becoming synonymous with knowledge graphs.
In previous years, the challenge was often to convey the benefits of semantic technologies. This has changed. The message has come through, and people get it now. Semantic technology and knowledge graphs became synonymous. The new challenge is how to deliver in practice.
To deliver, semantic technology must be developer-friendly. “Data usability” is important. That’s a weakness of the traditional RDF/OWL stack and perhaps the main thing that had been holding it back now.
In our stack, we are using a highly synergistic combination of RDF, SHACL, SPARQL and GraphQL. It provides simple-to-use access to the power of knowledge graphs for the mainstream developers. We are calling it Semantic GraphQL.
Nearly two years ago, we added GraphQL access to the enterprise knowledge graphs captured in TopBraid EDG. This was a new approach and it has proven to be extremely popular with our own developers and with our customers. We are now hearing that other vendors are starting to do something similar.
As more developers are using knowledge graphs to build solutions, the trend we see already happening and strongly continuing in 2020 is to provide an even more mainstream access and tools, ungeeking the use of semantic technology.
- Knowledge graphs will enable the governance architecture of the future.
We see knowledge graphs providing dynamically accessible, flexible and extensible business models for the enterprise with a trend toward enterprise knowledge graphs enabling a governance architecture of the future.
Currently, due to legacy IT environments, data governance is often an afterthought. It is an add-on layer retrofitted after data sources and applications are introduced in the enterprise. This “rear mirror view” approach means that connecting and capturing necessary information, at the right levels, is very difficult. Organizations spend a lot of money and time on implementing data governance solutions, but practical results are often modest. Yet, for all enterprises today, data is essential to their operations and for many data is their most valuable asset. Further, the complexity and variability of enterprise infrastructures is only growing. This means that data governance must become more of an operational reality within enterprises. To achieve it, engineers need to be able to build governance into their systems from the start just as they do with key aspects of security, access control, logging, and analytics.
Information that is necessary to manage enterprise data as an asset requires connecting different viewpoints and stakeholders – a natural fit for a knowledge graph. Knowledge graphs provide the basis for a forward-looking approach that data owners, data consumers, data stewards, architects, engineers, and all data stakeholders can adopt and utilize in their designs, development, and operations. With a knowledge graph approach, ‘to be governed’ assets can be dynamically modeled and the necessary metadata collected at any time as an ongoing part of any enterprise system (as it evolves from design to in-use) instead of only ‘after the system.’