Both personal knowledge management (PKM) systems and AI-based content generation tools are becoming increasingly popular today. How can we use them together?
Imagine, you have your ideas stored in your favorite PKM, such as Obsidian or Logseq (or the less favorite Roam Research one). How can you get insights about the knowledge you have?
You can use InfraNodus to import your PKM vault and visualize both the connections between the pages as well as the content of those pages as a network graph. You can then use the additional insights you get about your knowledge and feed the data from Obsidian to GPT-3 in order to develop your ideas further.
The advantage is that you're using advanced text mining and network analysis algorithms to provide much more insight into your knowledge graphs, rather than sending a simple prompt to GPT-3.
In this tutorial, we demonstrate the workflow that will help you do that. The main steps are:
1) Get an overview of your knowledge
2) Identify the main topics in your discourse
3) Identify the most influential pages / concepts
4) Discover the structural gaps within to bridge with the new ideas
5) Use the GPT3 AI in InfraNodus in combination with the analytics above to generate questions / ideas that would bridge those gaps and generate insight
Follow the steps below or watch this tutorial for an in-depth look:
Step 1: Import Your Knowledge Graph
Using your favorite software, Obsidian, Logseq, RoamResearch or RemNote, find the export function and export your whole vault (or a part of it, e.g. a folder) as an MD (markdown) archive.
Step 2: Upload the Content to InfraNodus
Go to https://infranodus.com/apps then select the Obsidian or RoamResearch app (you can also use it for LogSeq data),
...and choose the files you'd like to import:
Usability note: It's better to import a part of your vault first (e.g. a folder) to better understand the functionality (and to stay within the free trial's limits). If you like what you see, you can then import your whole graph.
Security note: If you don't want to store your data on InfraNodus servers, you can simply delete your graph after the analysis. We have to get it into our database to apply graph analysis algorithms. However, our servers are secure, they are based in Ireland, EU, and are in full compliance with GDPR. We don't ever have access to your data and, of course, we would never offer it to the third parties.
Step 3: Visualize the PKM Knowledge Graph
InfraNodus will then visualize your data as a network graph:
You can see that your dashboard consists of several elements:
- The Text Graph itself with 3 views:
- connections between the pages (based on your own backlinks)
- connections between the concepts (based on text network mining!)
- both pages and concepts
- The Analytics Panel, where you can find:
- the main topics (distinct color on the graph)
- the main concepts (bigger = more influential)
- structural gaps in your knowledge (based on network analysis of your discourse)
- + many more text mining measures
- AI insight panel, where you can:
- generate research questions, facts, and ideas based on the gaps or on the nodes you select in the graph
- Text statements panel, where you can view your original textual data
- Text editor, where you can add your ideas and save them into your knowledge graph
Step 4: Get an Overview: Identify the Main Topics and Key Concepts
The first step is to identify the key concepts and the main topics.
You can do so by looking at the graph: the bigger nodes are the pages / concepts with more influence. If they have the same color, they form topical clusters: i.e. they tend to appear together in your discourse.
For example, in our graph, these are:
1. "fractal dimension", "fractal", "consciousness" (light green)
2. "fractal scale free", "propagation", "power law" (purple)
3. "network "scale free" (green)
4. "small world", "topology" (pink)
This is a mix of the pages and concepts we use, but it gives us a clear idea of how our backlinks fit with the content we have.
Step 5: Filter the Concepts / Pages Graph View
Switch between the "Concepts Only" and "Pages Only" graph view:
The "Concepts Only" is a result of text mining done by InfraNodus, where it shows the main concepts you use in your texts and how they are connected.
The "Pages Only" is a visualization of the backlinks connections you have in your graphs.
The common view shows it all at once.
Note, when you change the view, the Analytics panel recalculates all your data, so you can operate at a zoom-out level (your backlinks) or at a zoom-in level (the concepts).
Milestone #1: By this time, you should have reached a good overview of your graph and the main ideas within.
Step 6: Identify the Structural Gaps and Bridge them with GPT3 AI
Now that you have an overview of your knowledge graph, you can also use it to generate new ideas.
The main difference here from the other tools is that the network algorithms that InfraNodus uses allow us to detect the structural gaps in your knowledge network. The structural gaps appear between different groups of concepts that could be better connected.
InfraNodus identifies the structural gap using the Gap Insight feature (in the Analytics panel below).
You can then use the built-in GPT3 AI to generate a research question / fact that will bridge that gap:
Step 7: Select Some Concepts to Link them Using the AI
You can also select some concepts directly on the graph (we recommend some that are not yet connected) and use the selection to generate the insight using the GPT3 AI:
Step 8: Add Some Of Your Ideas Into the Graph
Once you find an idea you like (or come up with it on your own with or without the AI help), you can save it into the graph.
Use the + Add Text > Interpret feature to save it under a different category "interpret" — this way you can then filter the concepts by that tag name and export them back into your favorite PKM:
Step 9: Hide Some "Obvious" Nodes from the Graph, to See What's Hiding Behind
To reiterate, select some obvious nodes on the graph and hide them, so you can see the concepts around them.
This helps you better see the interesting content at the periphery of the graph, not normally visible when you look at the main concepts:
Step 10: Export Your Ideas to Your Favorite PKM
Once you reiterate through steps 4 to 9, adding a few ideas into the graph, you will them them all tagged with the "interpret" tag.
The ideas that you automatically import from the AI will have the "ai" tag in them (e.g. "ai facts" or "ai questions").
You can then filter all this content in the statements view and export it into your favorite PKM system as a markdown (md) file, so you can integrate the ideas generated with InfraNodus into your existing discourse.
Try this workflow on https://infranodus.com and let us know if you have any questions, comments, or feedback!
Please sign in to leave a comment.