I'm a computer scientist and on this page I offer my thoughts and creations for the public to enjoy or to ignore.
Anki Incremental Reading
Have you ever looked anything up on Wikipedia? I'm sure you did, but do you still remember what you wanted to know at that time, or would you have to look it up again?
To give you an idea of what I'm talking about here, let me quickly tell you how incremental reading influenced me: In these days, I am learning a lot of material using it. The material is on all levels of expertise and the subjects include biology, chemistry, maths, history, geography, physics, astrophysics, computer science and so much more. With incremental reading I don't even feel overwhelmed, although I'm reading dozens of articles in parallel.
What is Incremental Reading?
Incremental reading (IR) has been invented by the developers of Supermemo. You can find a lot of material that describes the benefits of IR on that webpage. Let me just give you a quick appetizer here on how IR works.
- You find an article that is of interest to you.
- Instead of just reading it all in one go you add it to your IR software.
- The IR software schedules reviewing times for the article similar to a flashcard.
- After you read a paragraph or two, you decide what information in there is worthwhile for you to remember.
- You extract the paragraph and turn it into its own flashcard, possibly removing stuff you don't care about.
- The new flashcard is handled by your software again, such that you will see that same paragraph once more, sometime soon.
- Once you see it again, you trim it down even more. Possibly extract new flashcards.
- Of course, your IR software also shows you the main article again, for which you just repeat this behavior.
- Eventually, you cut down paragraphs into single chunks of vital information you want to remember. Through the constant trimming you ultimately end up with traditional flashcards, which can then be handled by Anki to make sure you do not forget them.
The Supermemo introduction to IR gives a small example of how this process works in practice.
IR with Anki
There's one big problem with IR: It's exclusively available for Supermemo. Try searching the net for "incremental reading", read up on it on Wikipedia - you'll always get to the same conclusion: Without buying Supermemo you are not going to get the benefits of IR. The concept of IR really convinced me, although I didn't even try it yet and I longed to be able to try it. As I have gained so much from Anki, I realized it was time for giving something back. I decided to give Anki users the option to try out IR for themselves and implemented an incremental reading plugin.
Installation of the IR plugin
Installation is as simple as it gets: Download the plugin. Open Anki, then click on "Settings>Plugins...>Open Plugin Folder" and copy the extracted files into that folder. If you're running on a Linux system, only copy the file AnkiIR.py and make sure to install python-libxml2 (f.ex. on Ubuntu: sudo apt-get install python-libxml2)
Restart Anki and enjoy - well, not yet I'm afraid. Even though you have everything installed, I suggest you also spend one or two more minutes to configure everything.
Configuration of the IR plugin
After you installed the plugin, verify that it is active (Settings>Plugins...>Startup>Incremental Reading should be checked). The plugin uses the normal scheduling of Anki, however, the IR process is vastly different from reviewing traditional flashcards. Hence, you should create a new deck dedicated to incremental reading only. Otherwise, you will mess up all your statistics - big time! For the new deck just stick to the Basic model, which has a "Front" and "Back" field. Do not rename these fields, as the plugin depends on these names. Additionally, I suggest you create another deck, into which you sort your final extracted flashcards for memorization. If you see, for example, 6 cards scheduled in your IR deck it makes a huge difference if these cards are articles or simple question-answer flashcards. If you separate the QA cards you get much more accurate information.
So what can you actually do with the IR plugin? If you click on Tools>Incremental Reading>Add URL for IR, you can load any webpage into Anki. For this copy a URL into your clipboard first, then click the menu entry to load the page into Anki (the QT forms have limited support for HTML. It works ok for most wikipedia articles, but may fail horribly on other webpages). After adding the card it gets scheduled as usual by Anki. When Anki shows you a card during your repetitions, you will get a new right-click menu. This menu has only two entries: Extract and Remove. "Extract" takes the currently selected part of your article and opens the Add Card dialog, where the content is inserted. This is the core of incremental reading which helps you trim down the articles to vital information. "Remove" simply removes the selected part, so you can throw out parts you no longer need.
It's up to you how you use the plugin. I just wanted to give people an opportunity to try out IR, but you'll have to find out which workflow works for you. Consider the following hints as a kind of starting point for your experiments.
- Contrary to normal Anki usage try to keep your new cards at zero. Remember that the scheduled cards are essentially articles you haven't read yet. There's no need to hurry with those. However, when you review one of them and start creating new excerpts and QA cards, they may get lost in all your other new cards. It works just fine, if you schedule new articles to the next day right after introducing them.
- After adding a wikipedia article, start reviewing it. If you have absolutely no time, then just schedule it as "Bad", i.e. it is shown again the next day. If you have a little time, then Extract the first paragraph. All wikipedia articles have this introductory paragraph that gives you an overview of the article. After adding the extracted summary schedule the remaining article a few days ahead.
- When you are shown a longer wikipedia article for review, simply Extract the first section (without even reading it), then reschedule the card. The newly extracted section will be smaller, and hence, easier to tackle. Make sure to schedule it sooner than the main article, so you know its contents before continueing the main article. The extracted section should be scheduled to the next day or if you have some more time you can already work through it.
- Once you broke down content to question-answer cards, give them a special tag like "qa". Every once in a while use Anki's export feature to export all of these cards and import them into their own deck. Then delete the exported qa-cards from your IR deck. After working with IR for a while you will have massively more qa-cards than articles you're reading and it pays off to have them separated for proper scheduling.
Following this "section extraction" scheme has a bunch of consequences of course: If you immediately schedule the extracted section as well, you can keep your scheduled/new counts down with next to no time consumption. The actual reading you perform will always be on small text portions, so you don't feel overwhelmed by large articles. Additionally, it is a great motivating factor, because every day your schedule will include small portions of all kinds of subjects. On the downside, processing a single wikipedia article in this way takes a long time - at least a week or two, for longer articles several weeks. But if you're bothered by that, then either you have not understood the concept of incremental reading or IR simply isn't for you. Don't forget, that you're not really going slow, because you fail to rush through a single article. Instead you can read dozens of articles at the same time. If you use IR and feel like you're not digesting enough new information, you're definitely doing something wrong.
Current version: 0.2.3
- minor fix for Linux
- more solid highlighting of extracted content
- automatic jumping to last extraction
- complete rewrite to update for Anki 1.2.x
- more solid parsing/processing (based on libxml2)
- extracted content no longer removed, but marked (green background)
- undo/redo support
- added hotkeys (r for remove, e for extract, c for extract+copy and shift+e or shift+c automatically adds the extracted card and closes the add dialog again)
- increased reschedule default to 5-7 days
- checking media database now possible (added workaround for bug in Anki)
- added Reschedule popup menu entry
- added undo/redo support
- initial release
- Some URLs cannot be added. (mostly, because they're not parseable)
- Color highlighting of extracted parts doesn't always work
- Hotkey support still broken (I hope to be able to add it again soonish)
- Not anywhere as powerful as Supermemo's incremental reading. But hey, it's free.
If you find anything else just let me know by e-mail or in the comments here. (Comments are moderated, so they are not shown immediately.) Of course, if you use and like the plugin, I also appreciate to hear from you.