2024 Explorations - Q1 Review

I started my 2024 Explorations in January. It was meant as a combination of two main objectives:

1) a learning framework or learning agenda, not so much to ensure that I would engage in continuous learning but more to ensure that my continuous learning was adequately focused on some key themes of interest;

2) using technology, and more specifically TiddlyMap, as a personal knowledge management tool that would allow me to experiment with (a form of) knowledge graph.

We've reached the end of the first quarter of 2024 and so far so good. I just completed a quarterly review of progress and generated a few insights.

• Is TiddlyMap allowing me to really learn about knowledge graphs? Yes, but as expected, it has its limitations. I will eventually crash the tool. I don't think it is meant as a graph database but it works well as an exploratory tool. Ultimately I need to move my data to a real knowledge graph tool like Neo4J. That should be a goal for the second quarterly. I started learning more about Neo4J, including learning the basics of Cypher.
  
  
• Is my learning framework working? The main themes and topics have proven very useful as guardrails and as an organization schema both for my thinking and for capturing notes. There are some issues with the taxonomy. Some topics are overlapping and I keep wanting to create more tags. So far, I have limited the number of additional tags and I have only made minor adjustments to the topic tags. Proliferation of tags would lead to inconsistencies. Until the tagging is automated, the number of tags is limited by my capacity to remember them all.
  
  The maps are telling me something pretty clear. I have focused perhaps 80% of my efforts on the AI and Knowledge Graph topics. The maps related to those topics are very large, which has enabled me to test filters. Like a search returning too many results, a map showing too many relationships is unreadable. For other topics, I have collected and curated resources, but I have not spent time connecting the dots. As a result, the maps are less interesting (so far).
  
  
  

• Is the basic ontology working? Yes, but the value of TiddlyMap's automated functionalities has made it much more powerful to create maps based on what TiddlyMap does with relationships based on tags and links associated with Tiddlers than to manually create a specific set of relationships based on my simple ontology. I have learned most by manipulating the filters to understand how relationships are displayed. Interpreting the resulting maps for potential insights is the next stage I want to dive into. Since I am the one creating all the links and the tags, the maps are not telling me anything I didn't already notice, but they are representing the connections visually and often reminding me of connections I made weeks ago that I don't hold in immediate memory.  I posted one of the maps in the Insight Maps section. 

The biggest ha-ha moment was related to tagging. I was using the tagging functionality to tag too many different types of things and failing to use a major functionality of the tool. I realized after a while that I should be using the fields to document the properties of a node. For example, "author" is a field rather than a tag. This allows me to have a consistent set of node properties and to rely on tags only for topics and some of the metadata used for navigation purposes. This also really helped make the maps more meaningful.

Using a GPT to get updates in topics of interest

About a month ago, I created a GPT based on ChatGPT 4.0.  It's easy to create but requires some fine-tuning.  I used Ross Dawson's approach detailed here:  Creating custom GPTs for news and Information Scanning; and I adjusted it to suit my own purpose.  The results have been mixed but I'm reasonably happy with what I got today. 

From Montaigne's "Essais" to Knowledge Graphs

Pretty much everything leads to a thought related to knowledge graph these days. Here is today's train of thought:

I was considering reacquainting myself with Montaigne's essays for a number of reasons.  

1. The style and how it relates (or not) to the blogging of today
2. The humanism/humanistic aspect of his writing and how it relates (or not) to today's conversations around humans and AI.  
3. His knowledge skepticism, introspection, questioning of his own knowledge, asking "Que Sais-je?"/What do I know?

 Digression Warning!

Montaigne was one of the authors I needed to study deeply in high school (French High School) to prepare for one of the end of high school exams.  In fact, the French Literature exam was not at the end of the last year of high school but at the end of the second-to-last year.  This involved very intense literary text analysis (for a 16-year-old) and an oral exam that required both presentation of a specific text and answering questions about the text from an examiner. You had to prepare a number of texts, come to the oral exam with a list, and the examiner would pick one and start drilling you.   I remember that our teacher preparing us for this exam was very demanding and therefore prepared us very thoroughly.  I bet that if by some miracle my list of prepared texts was put in front of me, I would suddenly remember a lot about each of them. Well, no great miracle needed. I found all my high school exams in the basement -- where all matters of interesting knowledge artifacts can be found.  I also have some of my handwritten (cursive), in-class philosophy exam essays, but I digress even within the digression, a sure sign that this should be a separate post. 

A couple of years later, I would find myself in English 101 in college in the US, totally lost trying to analyze Shakespeare and other English language literature not only because English was still challenging for me, but because the type of text analysis expected of students seemed so different.  I didn't "get" the assignment and struggled in English 101.  Perhaps this was an early lesson in how language, literature, and culture are so interconnected and part of what makes us so uniquely human.

End of Digression

I went down to my basement book collection and while I don't seem to have any Montaigne on hand, I did find a "Dictionnaire de Citations Francaises," 1978 edition. Luckily, quotes from long-deceased authors are reliably static, so this isn't a book that would age with time.  In fact, it's probably more accurate than most web-based collection of quotes.  I wanted to dig into some Montaigne quotes.  

There are multiple pages of Montaigne quotes, all from his "Essais". 

It's a heavy book, like most physical dictionaries, with a narrow page.  It's also a beautiful example of organized knowledge, with multiple indexes and numbered references.  I can search by topic, by author, by historical period.  So, immediately, I think... this needs to be turned into a knowledge graph.  I want to be able to visually SEE how these 16,460 quotations are connected.  Would it tell me something I can't possibly see by reading the dictionary?  I would think so. Perhaps I should try on a small scale.  

That being said, focusing on individual quotes extracting from essays could really fail to convey the context and full breadth of meaning and nuances that you would get from reading the full essays.  If I were asked to explain the meaning of a quote, wouldn't I want to know what was written before and after the specific quote?  So, while a knowledge graph based on individual quotes might be interesting as a small scale experiment, I can already see how it would have significant flaws, unless it could be paired with access to the full text for sensemaking purposes.