Sorry for the long wall of text, but please, help me.
I’d appreciate any feedback, but how can I implement the following logic in Anki for language learning?
I’m not a linguist, but I’ve noticed that any language follows this structure: it consists of sentences and constructions, then parts of speech (verbs, nouns, etc.), then word forms, and only afterward — several possible translations into the native language. Conceptually, it looks like this.
There are many methods for learning a language, but my brain personally chose the method of direct association: Word Form → Translation. At least at A1–A2 levels.
According to the Twenty Rules of Formulating Knowledge, an Anki card should contain one word form on the front and one translation on the back. Naturally, there should be two card types: direct and reverse translation.
Can all translation variants be placed on a single card? Yes, and I used to do that. But when a word has up to six translation options (I limited my model to six), the brain tends to remember only one — the one it’s already familiar with. Will I recall the fourth translation in a real-life situation? Of course not. So I started creating separate cards. Besides, putting all translations into one card violates the Twenty Rules.
As a result, in the general case, each word form may require up to 12 card types, and all of them will have the same frequency ranking, since they’re based on the same word form.
In German, typical verb forms include: the infinitive, 6 Präsens forms, 6 Präteritum forms, Partizip II (for Perfekt), and 2 imperative forms — 16 forms per verb.
Now let’s calculate: 16 forms × 6 translations × 2 directions = 192 cards per verb!
Let’s say we take 10 verbs — that gives us 160 forms, or 1,920 cards.
I think you’ll agree that studying all of them blindly would be overwhelming.
Okay, let’s narrow it down to infinitives only: 10 verbs × 6 translations × 2 directions = 120 cards.
But once I start studying cards like Infinitive → Translation_1, Infinitive → Translation_2, I immediately lose the link between the form and the meaning — I don’t remember which translation goes with which number. That’s why such cards can and should be combined into one: Infinitive → Translation_1–6.
This instantly reduces the number of cards from 120 to 70.
So now, for 10 infinitives we only have 70 cards, which should be studied sorted by frequency.
The ideal study order:
Translation_1 → Infinitive_1
Translation_2 → Infinitive_1
…
Translation_6 → Infinitive_6
only then: Infinitive_1 → Translation_1–6.
(This can be implemented via a subdeck “Infinitiv” with card order set by type and then by position.)
This all needs to work in combination with the burying mechanism, to avoid overwhelming the learner with all sibling cards in a single day.
Now let’s model a situation where each note contains only one word form with one translation, and only two card types: direct and reverse.
First problem: if the front side shows an infinitive — you don’t know which translation it’s asking for. Even if you write “Answer with translation #2 of 6”, you still have to remember what #2 means — this is unnecessary and distracting.
That’s why a card like Infinitive → Translation_1–6 is ideal.
If you split this into six separate notes, you lose grouping, and have to learn each pair individually — this increases the total number of cards.
Worse: these cards are no longer siblings — the burying mechanism won’t work.
And since they all have the same frequency rank, they can appear on the same day. This is bad for memory.
Moreover, the total number of cards becomes huge, and you lose control:
How many verbs (lemmas) have I studied? Where are the weak spots?
I tried linking cards using the Anki Note Linker plugin, but it only works visually — it doesn’t affect scheduling.
So for me, the ideal structure looks like this:
One main deck: German
Inside it — subdecks by part of speech (verbs, nouns, etc.) These subdecks aren’t for studying by part of speech, but to manage new card limits and statistics.
Each note represents one lemma, with multiple card types. Some cards have the same front side (e.g. the infinitive), but different backs — Translation_1–6. Yes, that’s 12 cards with the same frequency rank. They usually appear one after another. But burying helps spread them out, and other words enter the process.
The result is an excellent learning system. The only remaining task is to sort all words by frequency, so priority goes to really useful forms, not to uniform study of all 192 cards per verb.
Yes, the downside of this architecture is a lot of fields and many card types per note.
But if it helps learning — so be it. Excel + import are great for managing the complexity.
I described this using verbs, but the same applies to nouns. They can also have several translations, plus singular and plural forms. Unfortunately, German isn’t like English, where plural = just “s”. I still remember the day I didn’t recognize “die Häuser” = “the houses”, even though I’d known “das Haus” since childhood — just because the plural form is so different. And there are many such cases. So, forms need to be memorized separately.
Total, sorting by card type is the only correct solution.
Does anyone have recommendations on how to handle a situation like this?
How do you study a language while still following the 20 Rules of Formulating Knowledge, and successfully test yourself on just the specific piece of information during review?
Maybe I need to change the architecture? But tell me how.
Help me! I’m even ready to add 100,500 fields to each note and store each form with its own frequency value — if that’s what it takes to get correct sorting.
Is my architecture excessive? I don’t think so. At the very least, it’s possible to leave some fields and values empty, which means the corresponding cards simply won’t be generated. But that doesn’t mean this information can’t be added later if needed.