anki-wikidata

Easily create high-quality Anki cards with data from Wikidata!

Effective use of Anki is difficult and time-consuming. anki-wikidata helps you create high-quality Anki cards with a minimum of effort.

Workflow

Let's say you're learning about Pablo Picasso. To get started creating cards, first find the Wikidata entity identifier corresponding to Picasso:

anki-wikidata id "Pablo Picasso"

Q5593: Spanish painter and sculptor (1881–1973)
Q92627252: photograph by Man Ray 224564
Q7121722: song of the proto punk group The Modern Lovers
...

Now, check what kinds of cards anki-wikidata can create for this kind of entity (Q5593):

anki-wikidata list Q5593

birth-country:
- In what country was Pablo Picasso born?
- Spain
...

Create a deck file which includes this identifier:

anki-wikidata add --card birth-country deck.yml Q5593

This will create a file deck.yml like so:

entities:
- id: Q5593
  cards:
  - birth-country

You can then pass this file to anki-wikidata build:

anki-wikidata build deck.yml

This will create a deck out.apkg containing the card you saw earlier:

• Front: In what country was Pablo Picasso born?
• Back: Spain

Installation

git clone https://github.com/langston-barrett/anki-wikidata
cd anki-wikidata
pip install .
anki-wikidata --help

Or, run it without installing:

git clone https://github.com/langston-barrett/anki-wikidata
cd anki-wikidata
pip install -r requirements.txt
env PYTHONPATH=$PYTHONPATH:$PWD python3 anki_wikidata/main.py --help

Examples

• ./examples/nyc-boroughs.yml: A deck for recalling which borough each NYC neighborhood is in.
• ./examples/years-of-lyndon-johnson.yml: A deck for recalling The Years of Lyndon Johnson by Robert Caro.

Effective use of Anki with anki-wikidata

There are many guides on the effective use of spaced repetition software. This section discusses some common recommendations and how anki-wikidata helps to address them.

Use concise prompts

A prompt is the front side of a card. Using concise prompts makes review sessions shorter and makes it harder to "pattern match" (see below). The prompts in anki-wikidata are hand-written, and are constructed with brevity in mind.

Use atomic prompts

"Atomic" cards express the smallest meaningful portion of an idea. With a non- atomic card, you can waste time reviewing one part of a card that you remember well because you forgot another part of the card.

The data model of Wikidata consists of subject-predicate-object triples. In a sense, most of Wikidata consists of "atomic" facts, so it's easy to convert this data into "atomic" cards. Again, the hand-written prompts in anki-wikidata are written with atomicity in mind.

Use unambiguous prompts

Each prompt should have a single, unambiguously correct answer. "The USA" and "Florida, Missouri" are both correct answers to the question "Where was Mark Twain born?", so this is an ambiguous prompt. A better version might be "In which US state was Mark Twain born?" The only correct answer is "Missouri".

The hand-written prompts in anki-wikidata were constructed to be unambiguous.

Avoid pattern-matching

Cards with extraneous detail allow your brain to associate the details of a card with the answer, rather than truly remembering the answer. For example, if you have the cards "What was the date of birth of John Lewis?" and "When was José Martí born?", your brain might associate the phrase "What was the date of birth of" with the answer to the first, leading you to be unable to remember the fact in other contexts. It would be better to have two cards with nearly identical wording.

anki-wikidata solves this problem once and for all by using the exact same wording on every prompt that asks the same question.

Avoid orphans

TODO

Encode from multiple angles

TODO

Avoid non-facts

TODO

Development

Install development tools with pip install -r dev-requirements.txt. Use black and isort to format changes, and Mypy to typecheck them.