Skip to main content

KMWorld Conference

The annual KMWorld Conference returned as an in-person event this year for the first time since 2019. The conference ran November 7th – 10th and included the co-located conferences Enterprise Search & Discovery, Taxonomy Boot Camp, and Text Analytics Forum. 

The geographically distributed Synaptica team came together at the conference representing our products, giving presentations, and attending sessions. Being live and in-person once again allowed us to have face-to-face discussions with attendees, colleagues, and customers.

As in previous years (2020, 2019, 2018, and 2017), I’m covering some of the themes, topics, and trends as I observed them at the conference. 

Taxonomies & Ontologies

As I wrote in my 2020 blog, taxonomies have reached maturity in the industry. Fewer people approached me during the conference to ask about what a taxonomy is and, instead, are stating they are ready for taxonomy work in their organization. Moreover, the even more esoteric world of ontologies is becoming more commonplace in organizations regardless of their industry. What was once relegated to the realm of life sciences is now finding new audiences across many fields. Because of this, both organizations and vendors need to support more complex knowledge organization systems using taxonomies and ontologies.

Interestingly and ironically enough, the definitions for what is a taxonomy, an ontology, and a knowledge graph (more on those below) varied widely across the presentations I was able to attend. Moreover, sessions discussing the use of ontologies didn’t just appear where expected in Taxonomy Boot Camp but across the other co-located conference sessions as well. In other words, while the importance of taxonomies for knowledge management have long been known, the knowledge management audience is embracing the information science aspect of their work and want to learn more about it.

A common theme I heard was around the reusability and scaling of taxonomies and domain models. Perhaps audience members have seen taxonomy projects fail because they could not grow and scale to meet the wider needs of the organization. The answers to scaling touched both on the sustainability of the models through careful upfront planning and the flexibility more widely available technologies like graph databases are providing.

Many speakers spoke about the necessity of a proof of concept (PoC) project to prove the value of taxonomy work. A good PoC should be repeatable and scalable, not a single-use project with narrow scope within the organization. To address this, a PoC should focus on a narrow area of the organization to prove the methodology but always have as part of its goals the ability to grow taxonomies, add taxonomies and ontologies covering different domains, and meet the needs of multiple use cases. Part of this is the flexible model expansion and change afforded by graph databases. Careful planning should go into taxonomy and ontology modeling, but they should also be adaptable and allow for change to tackle the unforeseen and unexpected. For example, new product lines or mergers can result in domain expansions. Planning for this possibility will make these events less painful. Careful taxonomy and ontology modeling at the start is building for reuse across multiple use cases.

Taxonomies are certainly getting more attention and understanding, but so are ontologies. Many people in the industry are learning they can use ontologies to model their knowledge domains without necessarily embracing all of the complexity. For instance, many industries require rules to dictate how multiple taxonomies can interrelate, but they do not need advanced modeling using OWL. In these cases, using simpler SKOS models can be enough. On the flip side, industries outside of life sciences are also making more use of SPARQL to directly query graph databases from other systems and SHACL to define shapes and enforce validations. While still not widely adopted, the frequency with which these technologies are discussed makes it clear that speakers and audiences are making more complicated demands of their taxonomy and ontology use.

Knowledge Graphs, Again and Again

Knowledge graphs were still the buzz this year with an emphasis on using taxonomies and ontologies as the guiding structures. As with taxonomies and ontologies, knowledge graph presentations found their way into nearly all the conferences as enablers of knowledge management, information science, and search.

Additional evidence that knowledge graphs are making headway in organizations were the inclusion of several new search vendors in the Enterprise Solutions Showcase. Many of these vendors were already, or were working toward, including knowledge graph schemas underlying the search platform to deliver more relevant and personalized results.

A related topic brought up in presentations and in audience questions was the preparation necessary to plan knowledge graphs including conducting knowledge audits to understand the context and purpose of knowledge graph development. As with taxonomy PoCs, understanding the larger organizational information landscape while launching a narrow, focused PoC project to prove the value is essential to the success of knowledge graph development.

Change Management 

A common theme across the conference, change management addresses techniques which can help people adopt and adjust to new processes and tools. Resistance to new ways of working and platforms is commonplace, even when users are not happy with the current state.

In many sessions, the audience wanted to know how to create new processes and tools in the organization supporting knowledge management and information science. Adoption of these processes and tools, no matter how positive they are for workers and the larger organization, is frequently more challenging than identifying the problems and finding solutions for them.

Several speakers brought up the WIFM (what’s in it for me) approach, appealing directly to end users and their everyday work. While getting upper management buy-in can be difficult, it’s very often the case that end users know the flaws in the current ways of working but are afraid of what changes may mean for their daily work. Creating messaging directly addressing what each role gains from following a new process or adopting a new technology platform is a good way to effect change management and increase adoption. If end users can recognize tangible benefits, they are more likely to accept the change.

At the heart of all of the change management techniques suggested during the conference was communication. Frequent, clear communication is essential so people know what is happening, why it is happening, and especially when it is happening. Employees don’t like to feel ambushed by unannounced and sudden changes. They want to feel as if they are part of the process, especially if it is impacting their daily work. Creating clear communications, and particularly in conjunction with the communications team, lets employees know what to expect. Tailoring these communications to different groups will also address the WIFM. 

Artificial Intelligence & Machine Learning

As in past years, there were several talks dedicated to machine learning (ML) and artificial intelligence (AI). In my estimation, more of these talks focused on the pragmatic and employable use of machine learning rather than the promises which AI may deliver in the future. People are clearly more interested in presenting on and learning about ML successes.

The conversation often turned to data governance and supplying machine learning models with clean and consistent data with clear provenance. In addition, there was more discussion about how bias creeps into machine learning models and what we can do to prevent bias in data sets. Since many people in the industry expect to be using more ML in the future, their work to create processes and platforms to capture and provide well-curated data sets needs to start now. Hence, there appears to be more willingness on the behalf of organizations to invest in the up-front time and effort it takes to ensure downstream machine learning models have well-curated and accurate data.

All in all, spirits seemed high at KMWorld, lifted by the return to an in-person forum and the chance to discuss topics both old and new with returning and new conference attendees.