UPDATE: Please visit
https://github.com/stilelits/AnkiDeckGenerator for the script that I used to generate this deck, as well as more updated documentation on the deck itself.
-------
This is the most recent version of a concept I've been working on for some time, where I started out trying to create a deck to teach all vocab from the EDICT project that is marked with a "(P)". Since starting this project, I have learned that EDICT itself is derived from JMdict, and "P" means anything marked in JMdict as "news1", "ichi1", "spec1", "spec2", or "gai1". JMdict is also better maintained than EDICT, and I believe the entries are generally higher quality, so I am altering the title and contents of this deck to indicate that all the data is pulled from JMdict now, not EDICT. I take no credit (or blame) for the vocabulary selection itself, but I'm working on a good way to present these words as an Anki deck. My fundamental goal is to teach this common vocab in a methodical order that makes it easy to learn this huge set of words.
The solution I have come up with is to arrange words within the deck so that you first learn all the words that can be formed by Grade 1 kanji, then progress to Grade 2, and so on. I am aware that the grade levels are designed for the learning needs of children, not adults, but I think they actually work quite well for this method, because often the kanji for related concepts (e.g. north/south/east/west, family members, etc.) all show up within the same grade level, which means the related vocab for those concepts will all be learned together as well. Within each grade level, you start out learning words that only use a single new kanji (e.g. 所), before progressing to multi-kanji words (e.g. 台所, and later 春場所). This ensures that you are always building on a foundation of what you already know, which is critical to learning effectively, but it's a trait that I have never found in any other Anki deck, which is why I decided to create this one. (NOTE: The only time this approach doesn't quite work is with abbreviations...空母 shows up in Grade 2, but it's an abbreviation for 航空母艦, which doesn't show up until after grade 6. However, there are very few of these.)
My first efforts at making this deck simply included a card for each word from EDICT, arranged by kanji level, with the reading/meaning on the back. Although I was able to respond with readings/meanings when I read kanji in sentences, I quickly noticed that I was unable to recognize the same words when they were spoken aloud, and often I was unable to recognize individual kanji when I saw them outside of words. Therefore, this new deck contains three types of cards, tagged as either "Kanji_Card", "Vocab_Card", or "Reading_Card":
A Kanji_Card will show up immediately prior to the first vocab word that uses the kanji in question, and the back of the card contains all the vocab that can be formed up to the current grade level. Please do NOT attempt to memorize and respond with this entire list...in my opinion, as long as you can provide the first word in the list, it proves that you successfully recognized the kanji. The first card in the deck is "一 (kanji)", and as long as you recognize it as "いち", meaning "one", you should go ahead and mark yourself correct. All Kanji_Cards show "(kanji)" on the front, so that you know you are not looking at a Vocab_Card. This also prevents the issues Anki has with duplicate card fronts.
A Vocab_Card contains the actual word in its kanji form (if it has one) on the front. The back contains one or more readings/meanings that apply to that word. The second card in the deck is "一", and the back contains the two readings ひと and いち. If you tap on a reading, it will expand to show you the definition, and you can tap again to hide the definition. In order to consider a Vocab_Card correct, you should successfully respond with ALL readings AND definitions for each reading.
The third card in this deck is "いち", which is immediately recognizable as a Reading_Card because it's all in hiragana. The back for this card contains all the vocabulary that uses that reading, even if it has not been taught yet...for example, this card shows 一 meaning "one", but also 市 meaning "market", and 位置 meaning "position". These cards are tricky to judge...personally, I respond mentally with all the vocab that I know so far BEFORE showing the answer, and then I require myself to respond with the definitions for unlearned vocab after the words are shown. Please leave a comment if you have a better way to use these cards...I wish I could make these cards hide all the unlearned words, but I don't think there's a way to do that in Anki. If you don't think this type of card is necessary, you can simply remove everything tagged "Reading_Card" from the deck.
That's pretty much it! It's a large deck, but you should be able to progress through it fairly quickly...the first four cards are all just the word "one", after all, and there's a lot of repeated information and redundancy in here by design. I use this deck myself, and I add 100 new cards a day, which is an average of 50 new words and 6 new kanji per day, which sounds like a lot, but the structure of the deck means that most of those words are built out of the same kanji, or are closely related in subject matter, so it's not difficult to learn all of them in a single day...if you're already learning "meat" and "cow", it won't take you any effort at all to learn that "cow+meat=beef", and "fish+meat" will be equally obvious. Plus, if you can keep up that pace every day, you will complete the entire deck in just under one year!
NOTE: I almost forgot to mention it, but this deck randomizes the font for the FRONT of each card, and you can tap on the word itself to get a different random font, as many times as you like. It also randomizes whether the text is written horizontally or vertically, with a 50% chance of each, but this can be adjusted in the card template if you prefer only one or the other. The back is never randomized, but if you tap on the word there, it will take you to an internet dictionary entry for that word. Both of these features can be altered or removed through the card template.
ANOTHER NOTE: In September 2021, I lost my original deck creation script and had to recreate it from scratch, so I took the opportunity to add in several features that I'd always wanted...like the random decision between horizontal/vertical text, and also fine-tuning the kanji "alphabetization" scheme so that the numbers no longer appear in random order, etc. I also added a tag to every card that indicates the VERSION of JMdict (the exact date/time that the JMdict_e file was created) that was used to generate it, to make it easier to update the deck. Essentially, you can search for deck:"your::deck::here" -tag:JMDICT_2021_10_10_22_17_17, and it will show you only cards in the deck that are NOT present in the new version (i.e.words that are no longer marked as common in the current JMdict). It's not critically important, but if the goal is to learn common cards, I think it's reasonable to make that task slightly easier by trimming out cards that aren't considered common any more.