How I Read

2018-10-17

I don’t remember the exact sequence of events, but a couple of months ago, I became aware of:

For years, I’ve liked the idea of being able to easily mark passages in books I read and have them trivially resurfaced later, but Kindle’s locked-down ecosystem has made this difficult. When I came across Readwise, and read the other two articles above, I was suddenly motivated to build a better reading workflow of my own.

By and large, this post assumes that you’re familiar with the links above and by way of warning, it has the character of a written-once-read-never braindump.

According to Readwise, there are three fundamental steps in any good reading workflow: Capture, Review, and Integrate. Simon breaks this down differently, but the Readwise categorization resonates better with me. So here’s my reading workflow, phase-by-phase:

Overview

Capture: Read in Kindle, Instapaper, dead-tree books, or PDF. In all but the latter case, use Readwise to aggregate. In the case of PDF, a custom script rips annotations from PDFs in a special location in my iCloud Drive.

Review: For physical books, use OCR and paste into Readwise’s manual entry. Mass-export data from Readwise. Import this, as well as the PDF export, into Anki, with a special note type and template. Review these daily.

Integrate: When I encounter a highlight with information that I’d like to remember more explicitly, I use Anki’s “mark” feature. Periodically, I work through Marked items and convert them into fine-grained question-and-answer cards in Anki, then suspend the Highlight note.

Capture

In the Capture phase, I’m concerned with capturing highlights with the lowest overhead possible. Source by source:

Kindle: Just highlight.

Instapaper: If I’m reading something on the web longer than a few paragraphs, I generally send it to Instapaper and read it there, highlighting anything that I’d like to be reminded of.

PDF: I save these to a directory named “Highlighted PDFs” in my iCloud Drive, then (on iOS) open them from the “Files” app. Unlike iBooks on iOS, Files supports highlighting and annotating.

Physical Books: Generally I just draw a line in the margin next to the interesting part, then scribble a bit on the lower right corner of the right page. This makes flipping through to find all the marked up pages much easier later on.

Review

The meat of my “Review” phase is simply opening Anki every day and reading a group of highlights, presented at increasing intervals using Spaced Repetition, but there’s some menial labour I have to do when adding new highlights before I can do that. Again, this part is broken down by format.

Kindle: Readwise can import highlights from Kindle, but it requires me to actually open a browser window so that they can scrape the Kindle site using a browser plugin. I run this each time I’ve created new Kindle highlights, after which I run a script that downloads all of my Readwise highlights and imports them into Anki (more detail on this shortly).

I’ve been told that I can access these more cleanly through Goodreads’s API, but I haven’t had a chance to try this out yet.

Instapaper: Readwise automatically synchronizes Instapaper highlights, and my same script reimports them to Anki.

PDF: I wrote another script that extracts highlights from all of the PDFs in this special iCloud Drive directory that I save them to. The output of this script is also picked up by my Anki import script.

Physical Books: I use the Readwise “Freeform Input” feature, but I OCR the text rather than typing it, using an App called CamScanner. Once again, this is picked up by my Readwise-to-Anki script.

Custom Scripts

Basically, I have two (terrible, terrible quality) things going on:

  1. Rip highlights from PDFs;
  2. Download all highlights from Readwise, merge them with the PDF highlights, and import them into Anki.

There’s not much to say about the former script. I extract the document title and author from the PDF, and the text of each annotation (as well as any attached note), and dump them in JSON format.

The latter script uses the very-useful https://readwise.io/munger, which exports all of your highlights in one JSON file. It then dumps these fields in a tab-separated-value file, alongside a digest of a few fields, which I can use as a unique key to prevent duplication in case I change the note later on in Anki. I then import this file into Anki. Any cards with matching digest fields are ignored. If I ever want to “delete” a note, I suspend it instead of deleting it, to prevent my script from re-creating it.

Anki Configuration

I’ve created a Note type called “Highlight”, with the following fields:

I also have a “Highlights” deck. I enjoy reviewing these separately from the rest of my cards.

I made a few minor changes to the Study options for this deck:

The “Highlight” Note type has one Card type:

Front Template

<div class="book source">{{Source}}</div>
<div class="book author">{{Author}}</div>

<div class="book">{{Highlight}}</div>


<div class="book note">{{Note}}</div>

Styling

@import url("_style.anki.main.css");

(see _style.anki.main.css)

Back Template

{{FrontSide}}

Actual Review

Now that these scripts are “done”, it only takes a few seconds to vacuum up all my new highlights into Anki (except for physical books). The actual review process is simply reviewing all the due cards each day.

Integrate

When I encounter a note that I have no interest in seeing again, I suspend it.

When I encounter a note that I would like to remember specific details from, I mark it.

Periodically, I search for marked notes in Anki, and create cards from them in my default deck, with questions and answers (contrasted to my Highlight notes, where there is no knowledge test, just a requirement to read it again). If there’s no value at this point in keeping the highlight in addition to the new integrated knowledge, I suspend it.

Highlights will tend to stay relatively fresh when reviewed in this way, and knowledge integrated into Q&A-style cards becomes permanent. The underpinning of this system is a commitment to reviewing all due cards in Anki each day, which luckily I do.

Conclusion

So that’s a braindump of my reading system. I’ll probably revise this post as I continue to refine it, or if I feel like converting it into something less braindump-y. To recap the core ideas:


How I Actually Want This to Work

I’d love to have this all implemented as a series of importers in Perkeep, and a script to generate an Anki import file from the contents of my Perkeep instance. Fun project maybe?