The Ultimate Folder System: A quixotic journey to ACE

Knowledge management
21

Much better to me.
I find I constantly have to “translate” the ACE words into TIPS.
So thanks for the tip :smile:

2 Likes
22

TIPS?:thinking: Let me introduce my MATE framework :wink: MATE stands for:

  • Media (sources: books, articles, internet bookmarks, youtube videos, movies)
  • Atlas
  • Time
  • Efforts

Then I understood I still need Areas (aka Spaces), cos. I want e.g. my personal notes and my job notes to live in separate folders, so it became:

  • Media
  • Areas
    • Atlas - general knowledge, “default” area.
    • Job
    • Musician - my hobby area
    • Personal - household, medical records, etc
  • Time - daily notes here
  • Efforts

Then I understood I no longer need Efforts as separate folder, cos. Efforts are in their areas folders: Areas/Personal, Areas/Job, Areas/Musician (and I don’t want to mix them, it irritates a bit)

So instead I use #effort tag and status property (:fire:on, :recycle:ongoing, …) and I just have this dataview query in HOME note:

table without id status, rows.file.link as ""
from #effort 
group by status

so, MATE folder framework became just MAT after deleting the Efforts folder.

P.S. @nickmilo thank you for making Ideaverse demo vault publicly availabe, I adapted these great ideas for myself:

  • Atlas - great naming for general knowledge Area.
  • Efforts - instead of project without deadline, much better naming as well.
  • Logs (#map/log) - as addition to daily notes was a revelation for me, it’s much more convenient.
4 Likes
23

Thanks much for your insight, @Abhuva. I appreciate the idea of loading the entire vault into a vector database as embeddings and using a RAG approach to look up information, including “hidden” relations. Could you elaborate a bit more on your approach? Specifically:

  1. Could you describe the components of your (RAG-inspired) implementation? I.e. what vector database, retriever model for note or passage retrieval, and how did you integrate it all together?
  2. Did you utilize any open-source plugins or write your own implementation?
  3. What were the main challenges you faced, and how did you address them?
  4. For those looking to adopt a similar solution, what advice or best practices would you share?

I believe the technical audience would greatly appreciate learning from your experience. Thanks a ton in advance.

24

Hey Hello Milo, Loving the food for thought.
I can believe you’re getting up at 3am thinking about taxonomies and categorization as I’m suffering from the same bane…
Here is some feedback that might make you think from another angle :

  • The LATCH model is wrong in my mind, the STIR model does a great job on some aspects but I feel is still incomplete.

  • Problems with the LATCH model :

Alphabet is a hierarchy, so this one is useless
Location, Time and Category, in essence are the same thing. They are “properties” that can be tied to an entity, I think those concepts lead to an “error” in thinking because they are too broad.

  • What does location mean ? Where i found it ? Where it is now ? Where i want it to be ? The historical map of where it has been ? Is location a “place” in spaceTime ? but then what about abstract objects ? do we define abstract locations ?
  • What does time mean ? Time when it was created ? modified ? changed ? finished ? given ? categorized ?
  • What does category mean ? Let’s take fruits, do I categorize by color ? glycemic index value ? weight ? season of availability ?

I suppose you get the idea.

From the STIR model, I would have the same argument against space and time but I LOVE the ideas of “importance” and “relatedness”.

I try to see the issue from another angle :

  • The only reason to organize it retrieval
  • In the digital era: retrieval is done through search and queries first, then exploration in the result of a search or a query.

When doing a query you want 2 things :

  • Filter results
  • Order results

For me “Importance” is the notion of “Ordering” results and “Relatedness” is the notion of filtering results.

Now what does all that mean ?

“Organizing” things is the act of attributing “Properties” and “Relationships” to those things.
A “good” organization model allow for

  • Easy, intuitive and fast property and relationship attribution
  • Easy, intuitive and fast search and querying, allowing filtering (based on any property or relationship) and ordering (based on any property or relationship)

Now based on that a folder structure is ALWAYS a VERY BAD “organizing tool”. If you think about it its a tool allowing you to assign a unique property (“path”) to an object. And by dissecting that property (Calendar/Month/January 2024/2024-01-05) we’re trying to create an “Explorable query result”. Meaning that we’re trying to cram down a potential infinite amount of Entity properties or relationship into a unique Text String.

The only reason we keep talking about it is because it’s the most “available” tool in our current digital ecosystem. And in some cases it’s the ONLY tool we have to organize things.
Also this tool present the same “constraints” as the physical world, making it easier to understand because people can always “relate” to the physical world. (I’ve put that flashlight in my nighstand >> Flashlight.Object is in World/France/Home/Bedroom/Nighstand).

Now if we just accept that fact that the folder structure is a subpar organizing tool. Then it helps the reflection :

  • We could “think” about what an Ideal organizing tool would look like and dedicate a limited functionality scope to a folder system

Now since I believe that as an Ideal organizing tool “depends” on the type of entity we are managing (because it depends on those entities relevant properties and relationships) then an exercise to reflect on a universal organizing system is actually an Ontological exercise of defining what “properties” and “relationships” are the “most common” and shared by all types of “entities”.

There are multiple attempts in the ontology discipline to do such things.
But for a usable example I like the KR ontology by John F. Sowa (a quick research and the official website contains a lots of interesting concepts). (Note that i’m not an expert in that field, and Sowa model is critiqued by other experts. It’s just a nice example to help me think)

One very interesting concept is the distinction he maked between physical and Abstract

  • Physical is located in some place and occurs at some point in time
  • Abstract has no location in space and does not occur in time

by that sense, the STIR model doesn’t handle Abstract entities properly.

I know we could go down the rabbit hole of saying that we don’t “organise” abstract entities, we organise “notes” wich are “physical entities” talking about “abstract entities” but the issue is the same.

A good example from Sowa’s model is a “Schema”
( A physical continuant is an object, and an abstract continuant is a schema that may be used to characterize some object.).

Where would you “put” a Schema ?

Anyhow I have no definite answer to all those reflections.
Thanks a lot for the work on STIR and ACE, it’s always a blast to at least try those model.

Hope that I could spark some reflection.

3 Likes