# An academic Zettelkasten using Zettlr

## Introduction
Around about this time last year, I wrote a kind of "white paper" ("Whither Zettlr" https://hackmd.io/@ZyBG2fDcQ5yi0ckzqNysdw/Sk38hr8e5 ) about the possibilities for the Zettlr software. It provoked a certain amount of discussion, but it turned out to be a relatively rough mixture of thoughts and suggestions, many of which it was tricky to see how to implement, so both the use and the development of the Zettlr tool continued largely as before. Since then I have spent most of a year concentrating on experimentation and data gathering, followed by catching COVID-19 over the Christmas break, several weeks getting back to strength after that, then a few more weeks coaching and tutoring my daughter with her Masters work. My PhD had stalled, and something needed to change, so my wife and I had a discussion, with the result that I have blocked of some (relatively) uninterruptable time and I am now working much more "full time" on the research and the writing for my PhD, and have consequently got back into using Zettlr.

One of the reasons that I had slowed down on the PhD work was that my knowledge management process had become a burden. I was not really seeing much benefit from the time I put in to crunching through a seemingly-infinite heap of papers. I had often complained to my peers and supervisors that nowhere in my PhD process had anyone provided (or even linked to) any training on _how_ to read papers and manage knowledge. Conveniently, just what I needed (or part of it at least) turned up in the study materials for my daughters course. A link to a Harvard document titled "How to read a research paper" ( http://www.eecs.harvard.edu/~michaelm/postscripts/ReadPaper.pdf ). This began the journey which is becoming a surprisingly effective (to me, at least) way of dealing with my mountain of papers. Now I am happily chewing through between six and ten papers a day, and best of all, I am growing my interlinked Zettelkasten in a way which seems to actually make sense and even to help me to recall and connect things rather than feeling like just a place to dump stuff which I will never read again.

## Reading and taking notes
I read through the Harvard document, and some others I found with a bit of googling, and settled on the following process:

1. Read the title; look at the name of the journal; look at the names and affiliations of the authors.
2. Read the abstract very carefully. What is the context and purpose of the article? What do they claim?
3. Read the conclusion. Does it mention anything else not in the abstract?
4. Write a one or two sentence summary of the paper based on the abstract and conclusion as an instant reminder
5. Read the first and last paragraph of each section plus anything else which sticks out.
6. take notes of the key assumptions, arguments, methods, data, and results
7. Make separate notes of any limitations, biases, problems and inconsistencies you spot
8. Evaluate the quality and impact of the paper
9. Grab any useful quotes and references

As I began to use this process, I continued to refine it to suite my needs, and added a section for the application to my research and somewhere to put some general links to concepts (see the next section for more about concepts). I also added a section with a checklist of tasks to do on this paper including linking to concepts and following-up references. Right now I need these tasks as I have a lot of partial notes from my old approach which still need these steps to catch up with my current way of working. In the future I hope to follow the advice of Raul Pacheco-Vega who recommends never filing a paper until it has been processed.

I have set up this whole template as a _snippet_ in Zettlr, and can pre-fill a new document by picking _paper_ from the snippets colon menu.

## Organising the Zettelkasten

Since re-booting my literature reading using the above technique I feel much more confident about my ability to connect, recall and use the content of the papers I am reading. I have also realised that the way I was using Zettlr in the past was also getting in the way of my thinking and my writing, so I set out to completely reorganise things. Don't panic. I made sure to take several backups before embarking on this. At least that's something you can do with a digital Zettelkasten!

Up to now I have been using Zettlr as a kind of central notebook for everything in my life, in the fashion of Evernote (but as a free local application and repository) and that side of things has gone very well. I now find it much easier to lay my hands on all those odd notes, links, plans and ideas. But while struggling to make sense of my PhD research amongst all the other strands of my life I have come to the conclusion that this has positioned my research a long way from the sweet spot of the original paper Zettelkasten. One of the big clues was in the node graph generated by Zettlr. Around ninety percent of my documents remained orphaned, despite my efforts to connect them, and the end result was a few small clusters in a sea of singletons rather than a single linked knowledge base. I had a cluster of papers and concepts linked around my PhD thesis, and a few other clusters around things such as my plans for a fiction series, some research I did for a postgraduate teaching qualification a couple of years ago, and so on. However hard I looked at it, it still did not seem like the Zettelkasten described in all the excited web pages I had read.

I needed to do something drastic, so I duplicated my previous _Notes_ workspace to a new one named _Research_, and set about pruning as many of the unrelated unconnected documents as I could. Away went all my fiction writing and ideas. Away went all the lab notes from my experimentation. Away went all the computer configuration instructions, meeting notes and random scrapes from interesting web pages that I might want to look at later. When I finished I was left with just two folders. One named _Papers_, which contained all the boiled-down versions of my reading, both before and after I upgraded my note taking, and one named _Concepts_, which contained a load of documents named for common aspects which I had noticed in papers. these have names like _E-Waste_, _Short-term thinking_, _Power measurement_, _Education_, _Academic Reluctance_, and so on. During my reading I had treated these concepts as if they were tags, but with the crucial advantage that I could add my own definitions and notes to those documents, and also add links to other complementary concepts if required. Stripping back my Zettelkasten to this felt like a breath of fresh air, but there was still something missing.

In my original structure, I had separate folders for my writing projects, each containing miscellaneous notes as well as a papers folder of its own. The connection between the project and its papers was implicit in the folder structure. When I moved to the simplified, two folder, structure that connection was lost. I decided to make that connection explicit, and co-incidentally to allow easy re-use of resources for more than one project. The use of named documents as "smarter tags" had worked well with the mapping of papers to concepts, so I continued in that direction. I could have mixed the projects in with the concepts, but that felt a step too far, so I created another folder and named it _Projects_, creating one document in there for each writing project. I then added a link to an appropriate project whenever I encountered a concept which seemed like it might be useful. There was also one remaining step. As well as the dimensions of _Concept_ and _Project_ there was also the dimension of _People_. Just as with projects, people could be considered as just another type of concept, but they felt distinct enough to need their own folder.

So the final _Research_ Zettelkasten consists of four folders: _Papers_, _Concepts_, _Projects_ and _People_. Every document is linked. Every document is findable. Every document helps contribute to the growth of my research knowledge. Every time I read a new paper I create a document in the _Papers_ folder and summarise it according to the process in the previous section, linking and creating documents in _Concepts_ and _People_ as appropriate. When I think of a new project I create a document in Projects and add links to any of the documents in Concepts and People which seem appropriate. New knowledge grows organically and independently via the links to Concepts and People with no need to manually connect papers to projects.

Use of the knowledge base as it grows is mostly a matter of following links to find papers, extracting key points, then ordering those points into an argument. The important links are the _inbound_ links to documents in the _Concepts_ folder. This provides a way to automatically locate all the papers in the collection which are relevant to that concept.

Just as with the classic paper Zettelcasten, I use links for all my relationships, but I also use tags. The drawbacks of tags used for relationships were discussed in my original _Whither_ document, but there is another use, and for that they work very well. I use tags to indicate the _status_ of a document, something which is completely orthogonal to its knowledge relationships. If a document is incomplete, or new, or needs something done to it, I indicate that with a tag. Then I can use the Zettlr tag search facility to quickly find all documents which have that tag, and work my way though doing what needs to be done. When done, I remove the tag. The big win with this approach is that the status is stored entirely in the system, and if I am interrupted while doing something, the tags will clearly indicate what remains to be done when I get back. I call this approach "Tags for status, Links for relationship."

I have tried to do my writing in Zettlr, but it has never felt fully comfortable. Zettlr doesn't provide all the tools I need, and it is not easy to add them. These days I use Zettlr to gather and link my research but do my writing in Overleaf.

As for all those other documents which used to get in the way and slow me down. They are still there, in the original _Notes_ workspace. I removed all the papers, concepts, and people documents from that copy and it still serves the very useful purpose of a general document store with information accessible mainly by searching rather than linking. By keeping the two workspaces separate, I am able to have both open while I am working generally, but close the _Notes_ workspace when I need to navigate my research without being forced to step around all that unrelated stuff.

## Potential Improvements to Zettlr
Although Zettlr gets me most of the way toward a smooth process for knowledge acquisition, analysis, storage, and re-use, there are some speed-bumps and potholes along the way. Some of these issues are dealt with in a better way by other software, so the problems are not unsolvable.

Following _outbound_ links is generally easy, although it could be improved slightly. All my documents make use of yaml front-matter, which means that the hover preview which appears when moving the pointer over a reference to another document is filled mostly with the auto-created machine-readable front-matter rather than the important, human-created text. Skipping the front-matter when rendering these previews could make them much more useful in this case, and would make no difference in the cases where documents do not have front-matter

Following _inbound_ links is much more difficult, which is a shame, as this is where much of the inspiration happens. Noticing that several documents all link to one particular one can reveal knowledge which was probably not obvious when those links were created. Currently Zettlr provides an optional sidebar which shows a list of all other documents which link to this document, are linked from this document, or which share any tags. In any reasonably interlinked knowledge base, this results in a very full sidebar indeed, particularly as I use the "_tags for status, links for relationship_" approach. In most cases the sidebar ends up containing a large proportion of the entire knowledge base

Arguably the documents related just by tags do not need to be in this sidebar, as related documents can easily be found by ctrl-clicking any of the tags in the document and initiating a tag search. Also documents related only by outgoing links do not need to be in this sidebar as they are already listed in the text of this document. This leaves only the documents which are related by incoming links, which cannot easily be found any other way. There is another problem, though. The sidebar is always related to whichever document is currently on view. As soon as you click on one of the documents in the sidebar it is regenerated to contain the documents related to that one instead of where you started. This means that what should be a simple process of stepping through all the documents which link to a particular concept quickly becomes annoying and fiddly. None of the ways of achieving this work quite right.

It is possible to initiate a search by ctrl-clicking, just as for tags, but this runs a full-text search and therefore fills the search results with a block of text for every instance of that word or phrase, even when it is not used as a link. This is much less useful in this instance than a simple list of documents. It is possible to go back to the originating document after reading the linking document, but that means either finding the previous tab or using the back arrow, but that gets increasingly complicated if you happen to follow a subsequent link to another document. Worse, even when you _do_ get back to the original document it does not indicate which one you have just visited, and sets the scroll position back to the top of the list. For documents with a lot of incoming links just stepping through them can be a very frustrating process, particularly if they are "below the fold" in the sidebar.

I suggest that a better approach would be to use the search panel for this task, so that the selected search results remain on view until dismissed and can be stepped through easily regardless of where the exploration of ideas takes you. Ctrl-click is already used for full text search, which remains useful, so I propose adding some kind of search option to list all documents with links to the selected document. An example UI for this might be a second button next to the `Open` on the preview, and/or a specific option in the search form. Classic Wiki implementations provide an inbound link search by clicking on the automatically-generated page title, but that is not an option for Zettlr.

The next step after searching and finding documents which meet some search criteria would be to provide some way to step forward and back through the results. This feature is largely taken for granted in text search _within_ a document in all the software I have used, but is oddly missing in the Zettlr search panel. Moving back and forth through a set of search results seems to be the closest that Zettlr can come to selecting and working with a subset of physical cards as would be done in a paper Zettelkasten, at least without major software changes.

I hope that with these relatively minor changes, Zettlr could approach the real power of a paper Zettelkasten which lies in finding and following unexpected chains of knowledge.