SRS Decks

Hello dear Reader, and welcome back. Today I will write about the layout of my new Anki decks. The creation of these decks involved some heavy thinking. In a brainstorming session I asked myself the following questions. "What goes together, and what doesn't", "How should the structure look like", "What do I want to have in one deck", "What kind of information is really needed", "What is the aim of the deck", "How can i combine information and add it to one deck, instead of several different decks, with several different strutures."

My decks have been very simple in terms of layout and they had three or four fields at most. But every deck had different layouts. Deck models were something I didn’t want to touch. The single Japanese model was working so I saw no need to change anything. But the number of decks increased, what was two decks in the beginning, grew to four, six, and nine. Each containing vocabulary, sentences, kanji, production, and what have you. It got tedious to switch from one deck to another when reviews are due, not to talk about keeping track of the information added. But this was not the main reason for working on a new layout, to reduce the number of decks, and to have everything in a central place.

Of course it makes absolute sense to have different decks for different things. One for sentences, one for vocabulary, one for grammar, or whatever else it is that one wants to review. But if you have a deck containing only vocabulary, or only sentences, you will most likely also have different decks for them. The reason is that the facts can come from various sources, and they are best kept separate from one another. This can be good but it leads to two problems. In one deck you can have a word that exists in two of your other decks as well. And it makes no sense to review the same facts in different decks. Much less to keep track of the words or sentences you already have somewhere. And there is also another thing that can be overlooked. Say you learned the word 明るい, and one source tells you that it means bright, while the other says it is light, and a third tells you to remember it as cheerful. These are all translations that are absolutely valid for this one word. When you have different decks for different sources, you will most likely end up with having all these translations, and when you see them in one deck where it means "bright," while the translation you are expected is "light," or "cheerful," what are you doing? I bet you fail the card or rate it as being hard.

When all the information is added to a single deck, with tags that are corresponding to the source they are coming from, there is no need for micromanagement. You review the word and if you get a new definition for it, you are adding this translation to the card. If you have only one vocabulary deck, you would most likely don't add this information to the card for 明るい. The reason is that the card already exists, and the field shows up red in Anki, so this translation will be left out. Or think if you add random vocabulary from the internet, with say Rikaisan, which translation are you going to remember? All of them? One that fits the context you have taken it from?

It is even worse when it comes to sentences and their translations! You can have a sentence that presents the same information. Depending on your source, and there is a translation for the sentence, you will have several different translations in no time. And when you are looking up the information somewhere else, or translating sentences yourself, you get lost in translation. If you have only one deck for sentences in Japanese, and you want to add one, this isn't likely to happen. Because you have this one sentence, and you have this one translation. In different sentence decks, you can have all of the translations for only one sentence, do you see what this leads to? You are bound to waste your brain-power to mull the same fact over. Every word can have several different translations. Every sentence can be translated differently, depending on who does the translation, and you are asked to understand what you are working on.

After all this talk about pros and cons of having single decks versus multiple decks, it is about time to go into more detail about my first attempts to come up with a good deck structure, and the layout I decided to use.

In my first Test-deck I selected the Japanese Model and add several fields to it. The fields I wanted my cards to have in a deck in general: Vocabulary, Sentences, Translation for Words, Translations for sentence parts, Grammar Information, Japanese definitions, Notes containing general information, Particles, and Audio. Once the facts are added, the cards can be generatead, and spread to the two other deck models. One containing sentences, audio, translation and notes, one with vocabulary, translations and audio, and eventually one containing only grammar information, and example sentences.

+ When entering information all the information can go into one card and split to separate cards.
– This layout is prone to get overloaded and not fun to add facts.
– It leads to add more information to one card than there has to be.
– Information is randomly mixed together that doesn't go together.

The next test-deck was split into three models right from the start. One containing only Vocabulary, one for Sentences, and one for Grammar examples, in the Japanese deck model. This was looking a little better. It still contained all the fields mentioned above, but the information has to be added only to those fields, that go into one of the three models.

+ The information is spread out to separate deck models just by entering it to some fields.
– It does not matter if the information going to this deck comes from the same source, a textbook, a website containing something that seems worth adding, or already existing decks that get merged into this one deck, it leads to a mix of random facts that don't fit together.
– It is a nightmare to tag hundreds of cards when they are going to a single deck

Finally I found a promising deck-structure that really worked.

This screenshot contains the two basic models for my main deck. The first model is for sentences, the other for words, both containing different field-names.

Each of the two Basic models has its own model properties, and a set of fields belonging to them.

Some of the fields exist in both decks, so if i enter information in the Meaning, Expression, and Reading fields of this deck, the information will automatically be added to the Example sentences model. Those fields are only there to enter a word and generate readings not because i need them there.

This is the template for my vocabulary cards. The word in Kanji is on the front, Audio, Kana and translation are on the back. My sentence cards contain a field for Expression on the front, Meaning and Reading on the back along with Audio.

In both of the above screenshots you can see the difference between the input fields of the two Basic-Models. If i want to add items to one or the other, I just have to switch between them, so i can add both in one go in only those fields that i need. I’m not sure if i want to keep the grammar field, and the Notes field, but for now they are there.

The most important thing about this set-up is that it is easy to expand and refine both with different card templates. If i want my Japanese 例文 model to have a production template, i just add it, and hit generate cards. I can also take parts of the other models and use it in the Japanese 単語 Basic model as template. Most likely i will just add another Basic model for Japanese – Japanese cards, and use information from the existing decks. If i decide that I want to have the deck to contain definitions for words in Japanese, this goes into this model. And all I have to do is selecting a card and the rest is copy and paste. I can go all crazy but i don’t want to overdo it.

The philosophy behind this deck and its structure is to combine several ways of reviewing facts. The facts that are belonging together, will be reinforced, and there is always context. This layout also allows me to do several different things during reviews in one go. I can work on production while doing recognition work on my reading and pronunciation. If for instance a sentence card comes up that has Audio on the front I will write the sentence down, and when a vocabulary card or sentence comes up next, it is for recognition. The reviews demand from me to be able to process the information and to switch fast in thinking. Important about this is that different regions of the brain are activated, and activities such as listening, speaking, reading, and writing are involved.

My reviews consisted of only doing recognition in one deck for several minutes, listening and reading in another, and writing in the third. The adaption to the tasks in the decks was easy because I knew what comes up next. This new layout makes reviews more natural. The only difference between natural processing of language, it happens in a controlled setting, which allows for taking some time to think things over. But the rest of it is just like using the language naturally. Talking, while reading a paper, and listening to a radio. All happens at once or in the case of Anki, different things, one after the other. But it is always different, no certainty that the next card coming up is vocabulary, or prodcution, reading or listening. The only thing that remains the same is the tagging of cards belonging to a single source. In that respect the cards will remain separated, and there is no danger in having sentences from 飛躍 with those from 上級へのとびら, which i don’t own. I can still control what I want to review, by setting everything up accordingly. But the fact remains that everything goes into one deck. This way I hope to speed up the process of adding cards, the reviews once I get used to it, and my overall progress. The mining of textbooks for sentences and vocabulary, building up facts in a systematic fashion, instead of slow going, is the last reason for doing it. This also leaves more time for fun activities, like watching TV, reading on the web, or listening to Podcasts, to round of a learning session.

I hope that this article gives you an idea how to structure your own decks. If you wish to tell me your opinion, or just wishing to say thank you because it was helpful, please leave a comment.

Leave a Reply

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s