Principles for Metadata Minimalism

I despise metadata. I really do. The obligation that I feel to add it to my notes is friction, and it takes me away from what I want to be doing. It’s a cognitive burden.

What I want to figure out is how much Metadata is essential / necessary for a PKM.

I’ve reviewed a lot of note templates that others use, and there is so much metadata. I’m sure it’s useful for these people but it stresses me out and reminds me of my days doing data entry. I do wonder how much is metadata is being added for “just in-case” purposes.

I’ve been looking at Andy Matuschak’s notes and he seems weave metadata very organically into his notes via linking. But that could just be a presentable output where the backend is more convoluted.

Not here to put down on anyone who liberally uses Metadata but I do want to gather a collection of thoughts on the following:

  • What metadata do people find to have actual utility in their PKM? (eg. Note status I use often, that is essential. Date created on the other hand, not I’ll ever actually use this. I get why it could be useful but does anyone actually find use for this?)
  • How have you cut back on metadata or automated its entry? (I don’t think automated entry should give something a free pass to be included. It should still have a purpose and utility)
  • Does anyone have a way of working that allows for metadata to naturally emerge while writing notes rather than explicitly filling in metadata sections.

I’m currently about 150 notes deep into my own PKM and this is a pain point for me. Looking forward to peoples thoughts. Would love to synthesize this thread into principles for a Metadata Minimalist. Looking forward to hearing peoples thoughts and considered opinions.

12 Likes

Currently I only use the following:

  • note status, number 0-10 where 0 is just added my template and title, and 10 is basically publishable in principle
  • source, who came up with this idea/hypothesis or who wrote the book, had the quote
  • note type, is it an evergreen, a MOC, a list, what is it.
  • tags, and I am very liberal with tag use.

I’ve never used note status for anything, source has come up, especially for one concept where the source was a conversation with a friend. Not valuable, but it sparked some memories and then I found some other notes from that conversation, which I put into the main note. Note type I haven’t used either, but perhaps at one point it’ll be useful.

The rest is content: A section for links, a section for free form text, a summary, links to other notes or online content that might or might not be interesting to check, and random thoughts associated with the note. This I use for all my notes, no matter if it’s a book or an evergreen or recipe.

In principle though, only source is actually useful to me. What is your use of note status? To me it just creates a bad conscience for all the notes I should fix, books to read, entries to update, and so on.

2 Likes

It’s late so this came out a little stream-of-conscious-y, apologies in advance… In my defense I was just reading some of Lion Kimbro’s 2003 book How to Make a Complete Map of Every Thought You Think which introduced the idea of MOCs and it is absolutely roughly 100 pages of stream of consciousness blowout, so it affected me a bit I think. :slight_smile:

I’ll bite. Having used multiple note taking systems in the past I’ve coe to Obsidian along with a Zettelkasten-style approach at the same time. It’s given me a chance to radically rethink how I take notes, or rather to rethink how to develop a system that works around the way I want and need to take notes.

The fact is I’m a metadata minimalist in many ways. I barely use tags and the ones I do use are essentially only for note types, and of those there’s really only two that actually have any meaning: source-summary-in-progress for notes containing all my highlights/notes from a source, and source-summary-complete denoting that I’m done processing that source.

That’s really it as far as tags. (I technically have a couple others floating around from early on in my thrashing figuring out how to set this up, but they are pretty much meaningless)

And even with those they aren’t actually needed. What I found is that instead I have a different way of adding “metadata” to denote sources: I put them all in a sources folder, and I adopted a syntax I saw Nick Milo (I think) use where he prepends the title with S-source name.md. So now they are easily distinguished in links and in the file system. And I could even get rid of the -complete tag because when I’m finished processing a source I could just drag it into a completed folder.

Another piece of metadata that I apply without thinking is the date, because I use Zettelkasten IDs in (most of) my notes. (most = not sources, and a few other notes, but most use it) The reason I do this is to eliminate folders and just dump all my notes into a single folder, so I can avoid namespace collisions. Then I create MOCs when the need arises to create topic entry points into my notes.

For reference I’m a bit further than you, around 500 notes, and I have 7 MOCs. How do I know? Because I just searched for file:(MOC) and counted them. They are in various stages of disarray – the more I use them the more attention I pay to their layout/etc. I also have a “Home Note” that links to the MOCs along with having what are essentially several “proto-MOCs” directly in the home note – proto-MOC is just a fancy term for a bullet list with a heading and some sub-bullets that each link to a specific note somewhere, because I added the note and it wasn’t associated with anything else so I slapped a link to it in the home note and then started aggregating them into groups as that made sense. At some point as a group grows I will spin it out into its own MOC.

My whole focus is to do something very different from what I did in previous note taking systems and tools like Evernote and even wikis, where instead of focusing on the hierarchy and metadata I’m focusing on the process and using a few basic heuristics to guide me. For example, every note must be linked from somewhere not just link to somewhere. My focus is on fluidity wherever possible. I’m even extending this into work project tracking now and its pretty seamless.

The result for me so far is that Obsidian is starting to reflect the way I think and work, rather than me adapting to a rigid tool. And unlike tools like Evernote which expect you to come up with incredibly complex tagging schemes, here every time I try to add a tag and start using it then I fairly quickly get a physical eww feeling and back out of it.

Its absolutely arguable that I’ve shifted metadata into a different form i.e. the S- prefix for sources instead of a #source tag etc, which is true, but at least this way the metadata is visible in the filesystem at a glance as well, which provides some value to me if and when I need to depart Obsidian for something better later on.

One thing I do use in the front matter as a result of the Zettelkasten prefix is the aliases field. I’ve set the Note Refactor plugin to automatically create it for me when the note is created – I use the Note Refactor plugin heavily because it not only lets me auto-fill that field (the ZK plugin does not) but it also forces me to start my thought on another note and then refactor which creates the forward link my rule of thumb requires. Once the Note Refactor plugin creates the note I just quickly remove the ZK prefix from the alias field and then I can update the from note to use the alias in the link if I want, or if I’m not bothered I can just skip updating the from note entirely and maybe I’ll update it next time I happen across it.

I’m also considering using emojis in filenames to denote states for a select few notes like goals, based on an article I came across the other day. Not sure about that just yet but its an interesting idea, if used sparingly. I experimented with using emojis in tags as some like Bryan Jenks (I think) do but every time I quickly abandon it. I’ve also tried using emojis to denote note type in file name but also quickly abandon it. My reason is that I don’t make clean distinctions between “this is a seedling” vs “this is an evergreen note” – A NOTE IS A NOTE, some are long, some are short, some become hubs that build topic clusters around them and perhaps eventually become or lead to the creation of a MOC. But that’s an organic evolution rather than a top down decision. It also eliminates me from having to think “what type of note is this” – right now my note types are simple, its either a source, a MOC, or a permanent note, there’s no spectrum of permanent notes or states of permanent notes. I can’t be bothered with it. I am in the process of adding meeting notes into the mix but those really are just permanent notes because they are a record of discussions and decisions, and are tied by links to teams (and from team notes), and are spun out of a daily note where I take the meeting note and then use the note refactor plugin to create them.

BOTTOM LINE:

  • virtually no tags
  • “metadata” in filename, sparingly
  • constant fiddling and tweaking and messiness
  • Obsidian helps me bring order to the chaos so this weird thing I’m building feels right

EDIT I should add my reasons for avoiding tags are that my prior experience with tools that encouraged their use was that they break down horrendously over time. Tags are very weak association tools, just saying “this is of this type of topic” isn’t very helpful. I don’t even use backlinks very much in the traditional sense.

5 Likes

This is something I found as well, but couldn’t quite put my finger on that feeling.

To add to my above, I also use source, forgot about that. But its just a section at the bottom, not a tag. For a little bit I was making sources a [[link]] back to the source note which was named authorname2014 or similar, but quickly found that my graph was polluted by having TONS of notes all pointing back to these literature notes. It was just pure noise. So now the sources section is just a section listing sources, with an optional quote from the source below it similar to Andy’s notes. Or maybe its just a link to a wikipedia entry or a thread somewhere in a forum.

@davecan: when I came to Obsidian I was very excited to find it offers a graph view (initially only 1 view for all notes together in a vault), but when I got down to it, it was like I had ran into a brick wall: found no use.

Later on there was the possibility to do a graph per note: same experience for me.

There is even a “daily graph” channel on Discord, where people proudly present their graph. When I look at it, I think “great, so what?”

To me the graph shows the links and link directions between the notes: what does one do with that knowledge?

The only use I see for graphs is to identify unlinked/orphaned notes.

So, out of interest, what do you use graphs for?

1 Like

Already loving this thread.

@andresni - you’re right about status. I do get a weird pang of guilt having a bunch of stuff in “seedling” status. I might kill it. Just leave it as note types.

@davecan - I have also started to incorporate my metadata into naming conventions to some degree. I wrote another post here where I use the :question:in note names to denote an unanswered question notes and :grey_question:for answered question while also getting around the punctuation restrictions in file names. Metadata in the naming convention would also remove the the friction of going into a note and adding a template straight away.

I also don’t like the cluttered nature of sources being linked everywhere. I kind of like the idea of literature notes hovering around as orphans like an asteroid belt, not linked to anything.

@Klaas The graph is pretty. You feel like you’ve built something and it’s a visual index of a notes content. I like to click on the big nodes and see what it is. Utility is limited but it brings me a bit of joy to see it grow and to be able to explore it. I also write out my note titles as sentences, and if you take a path through a bunch of connected nodes just by hovering over them you take this cathartic journey through the graph that feels like a high intensity, stream of consciousness download of interconnected ideas that have been written down. I’m weird though.

3 Likes

Interesting discussion. But perhaps we can start with a definition of the meaning of ‘metadata’ in the context of this thread?

For instance, a note’s title is technically also metadata. But does this fall under the scenario painted in the opening post of ‘having so much metadata that it feels like data entry’? I suppose not because a note’s title is an organic piece of the note? :thinking:

Good question. Let’s not limit the definition then. Let’s say that metadata is anything that can be intentionally queried in a note. If some novel ideas result from a broad definition then all the better. I think many people will limit their thinking about metadata to fields that need to be filled out (I certainly did). Let’s shake that assumption up a bit.

I’m happy with my tagging system so I’ll share. I have a minimal set of metadata for the type of note indicated by tags:

#(my initials) - my own thoughts
#lit - literature
#moc - map of content
#more - I left the note before finishing something specific that I want to return to
#idea - marks a specific idea that I generate

These tags make it easy to filter out literature in searches / graph view. It takes zero brain space to decide what tag is appropriate.

Anything related to content is not tagged - it is only linked.

I do not care about the date I had an idea, and I do not use obsidian for task management so I guess that simplifies things.

I originally included metadata in titles, but realized this prevents obsidian from identifying unlinked mentions.

Hope this helps!

1 Like

Aliases solve this problem quite well. As @davecan mentioned above, coupled with Template and NoteRefactor plugins this helps you reduce the friction of adding aliases to almost imperceptible.

1 Like

@Klaas The main use I have for the graph at this time is for building out my notes by collecting notes from various other sources and integrating them into Obsidian. This involves a lot of bulk copy/paste and then combining notes into meaningful documents in Obsidian and setting up proper crosslinks. So for right now I have the graph configured to only show the links pointing to the currently selected note. This way as I compile my notes from these various external sources (other note taking tools, etc) into Obsidian I can see immediately if the note I just created is linked from one or more places or if its an orphan.

So I’m not using the graph as a tool for thinking and navigating, only to give me confidence that the new note is attached somewhere so I can throw it out of my mind and move on. I do get some value (or at least feel I do) from seeing the connected components that form on the global graph.

Oh on that note, you can adjust the settings under forces in the graph to change how the graph appears. When I did that it made the bridge notes more obvious, so I could see that a few core concepts connect multiple otherwise-disconnected islands of notes/ideas. That gives me some ideas on thinking about those more consciously in my daily life to see where else these core ideas can be applied – since they apply rather broadly they can be a useful jumping off point when learning other material, i.e. when learning if I look for these patterns in the new info then perhaps I can more easily grok the new material when I find them. Who knows, its just an idea. But maybe its one I should write a note about… :thinking:

@JAG interesting – I was looking at that same emoji for the exact same reason the other day! :smiley: But I hadn’t thought about using two different colors like that, what a great idea. I haven’t settled into a full naming convention that uses emoji but it seems like an interesting direction.

Agree also on the visual aesthetic of the graph. It’s not something that is incredibly useful all the time, but it does “spark joy” so I should keep it. :slight_smile:

Can you give some example note titles as sentences? I have a mix with a lot of topic/fact note titles where phrases don’t make sense, but for things like insights I have those do become phrases, much like Andy Matuschak. For example, in a set of notes extracted from some college notes I have a note titled Shannon's General Communication System which is just a diagram and a few short paragraphs describing the concept. I also have one called Signal characterization methods which just briefly describes how signals can be characterized in different ways, e.g. sync vs async, digital vs analog, etc. These are facts and concepts that don’t really have a rule applied, so they are what I would call reference notes (though that has connotations in ZK that I’m not applying here – I simply mean it is a knowledge base type item). But then I have my own insights I drew from notes which are written as sentences or phrases, like Discard information to gain clarity or Understand the observation scheme or Prefer links that connect knowledge over connecting notes.

I’m curious if you have the same variation in title types or if you force every title to be a phrase/sentence, and if so why.

Out of curiosity @mneinast do you use folders, or do you keep your notes all (or mostly) in the same folder? If in the same folder, how do you deal with filename collisions for the same term used in different contexts? This ultimately was why I opted to use the date-based IDs, not because they were dates (though that is a potentially useful side effect, maybe) but that they are easily ignorable unique IDs.

I think we have similar methods of using links. If I were to categorize my note types, they would be:

  • Concepts which are single words but I like to add an alias with the concept definition as well so I am getting a reminder of what the concept literally is every time I link it somewhere.
  • Notes which are named as a phrase eg. [[Aphorisms are both supportive and antithetical to philosophical thinking]]
  • Questions which are full sentence questions
  • MOCs

I like the titles of my notes to be sentences because they more organically flow into the context of another note which is nice and prevents me from being lazy and saying “also see: this not”.

I’m still very early days so I’m not speaking from a place of experience or authority.

I have one folder for each note type (personal, lit, moc) plus folders for templates and attachments.

Lit notes are automatically generated with a unique name using the Citations plugin. I guess these are marked with metadata: @

So far I’ve avoided conflicts in personal notes: maybe because the titles are phrases that I’m unlikely to replicate? I don’t make many one-word notes, except in rare cases where I decide to make a MOC.

I wasn’t aware that it is so easy to make aliases! That does seem like a good solution.

That’s interesting. Do you get particular value out of the note titles as questions? I haven’t created pure question notes yet. How do they fit into your system and workflows?

Again, it’s early days but so far I really like it. I’m taking the approach of. “Everything’s a MOC” (for a laugh, watch this comedy sketch and replace every mention of “drum” with MOC). Questions become their own mini maps of content, as are notes and concepts just through how they’re linked. Actual MOCs are still present and more informal in their structure.

Questions give direction, they are often involve multiple concepts and they naturally force you to link ideas together. It’s also really easy to go out and discuss a question with someone to get a quick new perspective. Not as easy with notes.

2 Likes

Re: Metadata in filenames…

I (at times, not always) use the method Karl Voit espouses (Managing Digital Files (e.g., Photographs) in Files and Folders) of including tags in the filename: append " – tag1 tag2" before the file extension.

It works well for me outside of Obsidian and I do it at times internal to the system.

I also have several home grown scripts I use to generate notes (in my Obsidian vault) from the command line and I use that " – " naming convention there also.

I find the discussion of dropping tags interesting and can see the value in places, but wanted to show a different way.

1 Like

It depends on the nature and goal of the note I take.

In my vault there are many kind of notes (a Zettelkasten-like implementation, task notes, MOC/Index/… notes, project notes, notes that model “things” (people, products,…) and so on. Some require many metadata (things notes are like database records), other less (in zettels are minimal).

The metadata I have in all notes are

  • created (date)
  • type (indicates the type of note discriminating if it’s in the context of zettelkasten/task/…)
  • domain (it is the more general subject the note belongs: it could be the project the note belongs, but also photography, developing)

In my zettelkasten notes I have a subject:: that is a list of subjects, in general I can have a up::[[parent]] if the note has a parent.

I have a minimal use of tags, I prefer having relationships between notes rather than a tag annotation. So, I prefer having subject:: [[macrophotography]] instead of #macrophotography.
One of the effects of this choice is that classification become part of the content itself. I can use [[macrophotography]] like a tag but remains a link.

The principle I apply is “the metadata I need for my goals, not one more, avoid over-engineering”.

  • If my goals will change, I’ll add other metadata :slight_smile:
  • if I discover that I never use the metadata X I added in the past, i just don’t add it anymore.

For example, I started two months ago to annotate my notes with tags #progress/seeding and #progress/evergreen, but I’ve noticed that I don’t use them, in the practice all my notes are always in a seeding-evergreen intermediate state (more than a simple draft, and never closed), so I’ve removed from my templates.

I think that the best thing for managing metadata in effective way is a careful design of templates of our notes. They capture most of all the effort.

1 Like

I use almost no metadata. Most of my notes have tags – in large part because I had a very large database of notes from before Obsidian that were already organized that way, so it made sense to me to carry it over. Apart from that, I use no metadata, ever, except where I need to add an alias/aliases for a note.

2 Likes

As I could see in my list, most entries from meta data came from „useful“ plugins to support special features. They are inserted by special templates not manually.