Taking Concept Mapping and Potentially Knowledge Graphs to Another Level

I can't pretend that I'm able to keep up with technology developments associated with AI but almost every week there is something that gets my attention. Today it's research from Meta into LLMs based on concepts rather than individual tokens.  

Large Concept Models: Language Modeling in a Sentence Representation Space, December 15, 2024. 

Link:  https://arxiv.org/abs/2412.08821

In the past year, I adjusted my mapping method to match what the technology could handle in terms of knowledge graphs, moving from multiple words and sometimes sentences to individual words, and paying much more attention to the connecting words. Now it is quite possible that the technology will evolve to mimic how we think and I can return to mapping in sentences rather than individual words. 

"In practice, a concept would often correspond to a sentence in a text document, or an equivalent speech utterance. We posit that a sentence is an appropriate unit to achieve language independence, in opposition to single words. This is in sharp contrast to current LLMs techniques which are heavily English centric and token based."

We don't think in individual words.  We think in combinations of words that express concepts. This assumes that concepts aren't individual words but rather snippets of sentences.  This is going back to the way I mapped conversations at NASA, using snippets rather than individual concepts.  This actually makes much more sense to me and always has. 

This research looks a language modeling in a sentence representation space.  I jumped the gun and immediately connected that to knowledge graphs and the challenges I've encountered trying to connect the dots around knowledge graphs and concept mapping. I'll have to dissect this and figure out if I'm totally off base. There may be subtle and not to subtle differences between concept mapping and concept or knowledge modeling that I am not grasping.

The "conceptual" approach described in the paper has potential implications for knowledge graphs:

• It could enhance semantic understanding:  By focusing on concepts rather than individual tokens, large concept models can better capture and represent the relationships and hierarchies in knowledge graphs.
• It could advance reasoning capabilities:  By operating at the level of concepts, large concept models could potentially perform more sophisticated reasoning based on the interconnected data in knowledge graphs, enhancing decision-making processes and inferencing. 
• It could also improve traceability and our understanding of how the models operate. Conceptual representation would be easier to follow just like a concept map made of sentences is much easier to follow than a concept map made up of words and connecting sentences. 

I'm sure there is more to it but that's as much as I can grasp right now. 

Pivots and Pathways: Learning, Unlearning, and Recombining for Growth

Imagine yourself in a room with just one door—the door you entered through. At first, the room feels full of promise, a new space to explore. But as time passes, you’ve seen everything it has to offer. The room becomes familiar, even stifling. There is so much more you want to experience, but the only door leads back to where you started. What you need are more doors, new ways to move forward. And here’s the remarkable thing: these doors don’t already exist. They are created through your creativity, imagination, and willingness to pivot.

A pivot is the hinge on those doors, enabling movement into new rooms, new pathways. Unlike a labyrinth with a single correct exit, these pathways are yours to design. Sometimes, a pivot starts as a window, giving you a glimpse of what lies beyond. When you’re ready—and when you’ve overcome the fear of stepping into the unknown—that window transforms into a door. The pivot is what makes that door’s opening possible, turning an idea or aspiration into a tangible opportunity.

Not all pivots are intentional. Life throws its share of curveballs, forcing us to adapt. But whether a pivot is chosen or imposed, it requires resilience and openness to learning. Navigating these transitions involves more than simply stepping through the door; it’s about learning through change, unlearning old certainties, and recombining past experiences into something new.

Learning and Unlearning: The Twin Pillars of Growth

Pivots compel us to learn new skills, adopt new perspectives, and stretch beyond our comfort zones. But they also force us to confront what we need to unlearn. In organizations, this is often referred to as shedding “zombie knowledge”—outdated practices that linger even when they no longer serve their purpose. The same concept applies to individuals. To embrace a pivot, we must be willing to let go of habits, beliefs, or frameworks that no longer align with where we’re headed.

Unlearning creates space for fresh insights and innovation. It’s not always easy—after all, certainty feels safe—but it’s essential for growth. New pathways challenge us to rethink our mental frameworks, making us more adaptable and curious. They push us to ask, “What else is possible?”

Recombining: The Creative Potential of Pivots

Of all the aspects of pivots, recombining is perhaps the most exciting. It’s the process of weaving together threads from your past to create something entirely new. Recent experiences, long-held interests, and latent skills can intersect in surprising ways, opening doors you hadn’t imagined.  I always find mapping to be a very valuable "recombination" tool, allowing me to connect previously disconnected ideas.  That's partly why I call it "insight mapping". 

For example, I’m currently exploring how to recombine my experiences with volunteering with older adults with my decades-long interest in Personal Knowledge Management (PKM). PKM has been a personal tool for organizing and making sense of information, but I hadn’t thought about applying it more broadly in my work until now. By combining these threads, I see the potential to develop services that help older adults navigate the complexities of later life, what Susan Wilner Golden describes as the 4th and 5th quarters of life (in her book Stage (Not Age)). This pivot has sparked a burst of creative energy, a new way of looking at how I can contribute.

Resilience and the Courage to Pivot

Whether a pivot is self-initiated or forced by circumstance, resilience is the bedrock that supports it. Resilience is not just about bouncing back; it’s about bouncing forward—using the lessons of the past to propel yourself into new territory. Each pivot is an opportunity to reassess what matters, to experiment, and to move closer to a life of coherence and purpose.

Conclusion: Crafting Your Pathways

Pivots are not detours; they are pathways. They challenge us to learn, unlearn, and recombine in ways that create new opportunities and greater clarity. The next time you find yourself in a room with no apparent exits, remember: the doors don’t have to exist yet. Your imagination, experiences, and aspirations can build them. And when you’re ready to step through, visualize the hinges, create your pivot and step walk forward.

What pivots are you considering, and how might they lead to new pathways in your life?

As a side note, good Personal Knowledge Management habits are helpful when considering and creating pivots.

Exploring Knowledge and Resilience in the Renaissance Quarter of Life

When Dr. Madelyn Blair invited me to her Unlocked show to discuss knowledge and resilience, it felt like the perfect opportunity to reflect on the themes that have shaped my past—and will continue to guide my future.

As someone who rarely steps into the podcasting world—my only prior experience was with Zach Wahl on the Enterprise Knowledge podcast—it felt fitting to reflect on the career pivots and personal growth that have defined my 30-year journey. Yet, it's the next 30 years that truly excite me.

Having left my full-time job in May, I am in a transition phase, shaping this next chapter to align with my evolving goals and a 30-year vision. One recurring theme in both podcast conversations has been Personal Knowledge Management (PKM). It’s a concept that has anchored much of my work and learning, and I see it playing an even bigger role in this new chapter.

Susan Wilner Golden's Stage (Not Age) has been an insightful guide during this phase of my life. In her '5Q Life Framework,' I find myself squarely in Q3, the Renaissance quarter—a time for repurposing, rejuvenation, and reevaluation. Golden describes it as the 'portfolio years,' a period of meaningful activities that blend paid and unpaid work. This resonates deeply with the title I’ve embraced for myself: Knowledge Explorer.

The Renaissance quarter is also an opportunity to leverage the habits and practices that have built my resilience over the years—while strengthening them for the future. It’s about intentionality, setting goals, and nurturing a vision that allows me to meet challenges with confidence. Resilience, like muscle mass, requires maintenance and care.

As I navigate this phase, I’m reminded that resilience is built through small, deliberate steps. What habits or practices are you leveraging to build your own resilience for the future?

Back to "School": Learning the Basics of Protege, Python, and Cypher

 A few years back, I inherited a substantial collection of vinyl records that accumulated over two generations and two sides of the family.  The result is a very eclectic collection that goes from field recordings of birds of Kenya to early career Johnny Cash and lots of Edith Piaf. I have a simple record player and once in a while over the past few years, I have played some of my favorite records.  About a month or so ago, having some free time, I decided to catalog the collection using my favorite tool, TiddlyWiki.  In the process, I learned quickly how to leverage some of the functionalities I had previously neglected in TiddlyWiki.  I had always used the tagging function but I had not used the metadata fields.

I now have a complete catalog with close to 700 records and this turns out to be an interesting data set for learning purposes because:  

a) I can save the TiddlyWiki as a CSV file with all the metadata.

b) I am at the point in my exploration of Python where I can understand basic instructions and use Python in Excel to manipulate the data, including (perhaps) the less structured data (all the album titles and song titles). 

I was getting bored with Python exercises that use meaningless data.  Hopefully using my own dataset will be a little more exciting.  Using the TiddlyWiki Records Catalog will also lead to some data cleaning issues.  For example, I'm not sure I can easily extract a list of songs even though I've "captured" all the data in the text section of each record. 

At times, when I am struggling to understand what I am doing with Python, I ask myself why I am trying to learn Python in the first place since I don't plan to become a programmer. I am reminded of my long-term goal: I need a better, broad understanding of data management and associated tools to be able to converse intelligently around the technical aspects of knowledge graphs.  As a result, I am learning the basics of Protege to understand ontology design, the basics of Python, the basics of Cypher to understand graph querying, etc..