Flash cards get a bad rap sometimes. Put a word or sentence on the front, the meaning on the back, and you’re done. Repeat this a few hundred or thousand times and you have a deck. Such a simple and bare tool. They’re just flash cards, right? Really…?
I thought it was about time I showed behind the scenes what goes into Jalup flash cards, and how they are far from just “flash cards.” You might even get your own ideas how to build decks in Japanese or another language.
Step 1: Research
Depending on the level of Japanese, the research requirements have varied greatly. Over time, there have been three main types of “Jalup research.”
Low Level: Jalup Beginner
Starting from scratch, I thought it would be easy to figure out the necessary basics to teach Japanese. But the order of how you introduce things is a complex subject. The good news is that there is a fairly similar approach among most textbooks, because it is logical. So I pored over a dozen textbooks (including a few English ones), to look for patterns. I agreed with some parts, others not so much, and put together an approach I was satisfied with.
Flash cards are good for small bits of information – textbooks love the opposite. This wasn’t going to work, so I had to get the essence of grammar down to the minimal clear wording (which took many tries to accomplish).
Since there is so much “beginner vocabulary” in most textbooks, I had to decide what was urgent and what was not. I wanted all of Jalup Beginner to fall within 1,000 cards. That meant I had to cut certain things. I could either teach all the numbers, days, months, colors, etc. at once or I could spread it out over time. I chose the latter because once you get one piece of a set of something, it is easier to teach the rest of the set later.
Medium Levels: Jalup Intermediate to Jalup Hero
Jalup Beginner was originally designed with the idea in mind that it would set you up to create your own J-J cards immediately after finishing. But when people were constantly telling me the trouble and failures they were having with J-J, I tried to find out if it was possible to do anything about this.
With my own studying years prior, I had amassed extensive experience creating my own Japanese only cards. Unfortunately, it didn’t help much, because creating J-J cards for someone else was an entirely different beast. I still didn’t even know at the time if it was possible. My own J-J cards relied heavily on my self-knowledge. I knew what I did and didn’t know. But to create a deck, I would have to expand the limited “knowledge pool” from Jalup Beginner without any outside assistance.
I had no idea where to start so I just dove right in with endless trial and error. It wasn’t as simple as just picking up a book and then adding Japanese words and definitions. Sentences needed to contain easy words and explanations. The two don’t necessarily go together. Cat is an easy word. Explaining it isn’t.
High Levels: Jalup Master, Jalup Champion, and Beyond
I made the decision to target specific books (manga) from this stage forward. I wanted to give people more confidence on specific media they could enjoy as well as provide level appropriate, fun recommendations for them.
I ended up reading the first several pages of dozens of manga to make sure that the level was about right, the language was actually frequently used, and whether it would be possible to follow from the current level of deck.
Decks up to this point allowed me to choose and skip whatever I wanted. I could save the trickier stuff for later. From here on out I had to include nearly everything.
Step 2. Find a word to introduce
Since every word relies on your previous knowledge, Jalup Beginner should have been the easiest. You knew zero words and I had the entire language at my disposal. But from the start I had to find just the right word at just the right difficulty, with just the right importance. Long term planning for flash cards was a priority. It is really more like a textbook split into linked pieces rather than just 1,000 separate pieces of info called a deck.
You’d think that the more cards and words that are made, the easier it would be to find new words. While I’d like to be able to say I’ve memorized all the words used in all of the Jalup decks, I haven’t. Which means whenever I go searching for a new word to introduce, I first have to search to see if I’ve already done it before. The further along I am in the decks, the more likely I can’t use a word I think is new.
What makes things worse is that I’m not just checking whether I can or can’t use a new word to introduce. I’m also checking all the words in the rest of the sentence, the definition and the note. In other words, i have to remember both what you do and don’t know, at every single step.
Step 3: Create an original sentence
Let’s say I wanted to introduce the word “red.” Simple. “This is a red pen.” Done?
While I’m not saying all the sentences in Jalup are interesting (plenty early on are run of the mill), I spent a lot of time making sure the sentences were memorable. Silly, stupid, weird, and emotional. Something that actually brings the sentence to life. A living sentence is one that you’ll remember. If a flash card has ever made you smile (or cringe), that wasn’t just luck.
A flash card isn’t just a boring piece of knowledge. It’s a story, wrapped up in one sentence. And in case you haven’t noticed, sometimes I tie more than one card together. You’ll see repeating characters and themes throughout the decks.
Step 4: Create an original definition
A good definition needs a perfect balance of information and ease. You should be able to look at it and think “okay, I get it.” And that’s just for Jalup Beginner.
Creating a Japanese-only definition from limited words is several times harder than creating the sentence on the front of a card. All current native Japanese dictionaries use too many words you don’t know (they expect you to be native after all) and often circle around back to the same synonym, resulting in a vicious cycle of confusion.
At first I tried just replacing unknown words with known ones. That didn’t work (until much later). I soon realized that I was going to have to create definitions from scratch. One positive side effect is that it made the definitions themselves more interesting, because sometimes a little mini story was required to get to the meaning.
Step 5: Add a note
Sometimes a definition isn’t enough and a bonus explanation is needed. Maybe an additional sample sentence. Maybe how the word is compared to another similar or different word. Maybe a point how the word may be used in different situations in the future.
Or maybe some oversight I realized later on, and this was the only place I could put an important piece of info without changing the order of all the cards…
Step 6: Edit. Edit. EDIT!
While it feels great to finish 1,000 cards, this leaves me far from finished. That moment was just a small victory before having to go back and fix the countless typos and errors. Card by card, I check:
- Spelling, grammar, spacing and other aesthetic issues.
- Comprehension (does this make sense and convey exactly what I’m trying to say?)
- Note necessity – does this card need a little extra boost?
- If I’ve used any words outside of what I was limited to.
Step 7: Edit – Native Style
The next round of edits come from a native Japanese speaker. For years now, my wife Yuki has been doing them (in the past I’ve also gotten help from others).
Card by card, everything is read in depth to make sure it all sounds natural. This was much harder in the beginning (especially Jalup Beginner) because I would get a healthy dose of “it would sound better if you change it to this.” The “this” would always involve some word that wasn’t introduced yet.
The limited vocabulary set was a challenge for Yuki as well (especially for J-J definitions). Natural Japanese without the usual freedom of choice requires a lot more work.
Step 8: Add in the furigana
Furigana, or the Japanese pronunciation guide that appears above kanji, is absolutely vital to every card, yet is so prone to mistakes.
I started off with an automatic generator, which uses a collective database to set up all the furigana readings. Thankfully this gets around 85% accuracy. Which is good… but also completely unacceptable. So I need to go through and fix all the problematic readings.
After I do this, Yuki briefly checks to make sure I didn’t miss anything.
Step 9: Package the deck
Once the deck is “complete,” I have to put it together for sale, set it up with the server, write up a long introduction post and actually get it out there so people can use it.
Step 10: Organize user-discovered card mistakes
Thought all that double editing caught everything? I wish. 1,000 flash cards is the equivalent of 2 full textbooks. There’s so much that can go wrong. Luckily, people on Jalup are kind, so when they discover errors, they not only point them out, they understand the inevitability of them.
This takes place over several months, as it requires many motivated users who not only finish the decks, but also want to help. I keep a running list of errors and once it gets big enough I add them all in at once. It would be impossible to update after every individual error as the update process takes time.
Step 11: Answer user questions
Despite my best attempts to be as clear as possible, I expect e-mail questions over the coming months about certain cards or issues a user is having while learning. I keep notes of issue patterns, and use this when I come around for later edits.
Step 12: Add audio
The original audio of Jalup Beginner several years ago was pretty bad. It was done using an internal laptop microphone, without much thought or editing. I later realized that good audio can make or break a learning resource and is just as important as all the other information provided on a card.
It needs to be clearly read, in a way that includes emotion (and isn’t just a boring textbook read sentence). Just like the sentences themselves, they need an infusion of “life” to make them feel more real. To record audio you enjoy and want to hear, Yuki:
- Uses professional equipment in a quiet home studio.
- Performs a lengthy voice check and warm-up (for every short recording session).
- Reads one sentence to herself a few times, figuring out how she wants to say it, and the way to make it sound best.
- Records one sentence several times.
- Listens back and chooses the recording she liked best. Occasionally she likes none of them and will record the same sentence again.
- Repeats the above 1,000 times.
- Discovers new deck errors which get added to the revision list.
- Realizes some sentences weren’t as natural as she originally thought and it was only reading them out loud that showed her that.
- Organizes all the files making sound edits for optimal quality.
I then listen to all 1,000 audio files, comparing them against the the cards. I keep a running tab of any cards that sound off, have background noise, or are wrong. Yuki re-records these cards. I check the re-recorded cards.
Step 13: Organize user-discovered audio mistakes
A repeat of Step 10, except with audio. Audio mistakes will be discovered and are listed and added to a later big update.
Step 14: Card linking
The least enjoyable part of the process, that takes forever and gives me countless headaches.
In order to get the card linking effect of the Jalup App, I have a large spreadsheet of all the sentences that I code manually, word by word. Every single card gets assigned a keyword, and then for every sentence/definition/note, keywords are added to all parts of a card. This means for one J-J Card, there are usually at least 25 keywords added. Times this by 1,000 cards and that is 25,000 manual keywords coded for one deck level.
I’ve gotten this a lot – “why don’t you just automate the process?” It’s been tried by myself and 2 other programmers (with way more experience than myself), and it just didn’t work.
Once I finish the spreadsheet juggernaut, I have to program the decks into both iOS and Android apps (which are built in separate languages) to allow for the app to handle the new deck in every function that uses it. Endless testing ensues.
I run an automated error check on the code linking spreadsheet I made. It’s highly character sensitive (intentionally), so one wrong parentheses or dash will cause an error to come up. I use all of these errors (usually several hundred) and go back and fix the spreadsheet accordingly. Yuki does a final check of the app, going through all the links. She attempts to tap every one of those 25,000 links…
The card linking process also points out ways I can update the older decks to make them better. Jalup Beginner and Intermediate were significantly improved when linked and uploaded to the app. Many notes were added, words changed, and clarifications were made. Advanced and Expert was also changed (though in more minor ways). For the people that complain Jalup Intermediate is too hard, you should have seen it before the updates, and especially before the card linking.
Step 15: Organize user-discovered card linking mistakes
Notice a pattern here? As with cards and audio, I get mistake reports over the coming months, and keep an ongoing list to update later. For each level, there are over 25,000 places I could have made a linking error. I do my best to keep this as low as possible.
Step 16: Repeat!
Everything I do here for a deck of 1,000 cards is repeated for each new deck. This is why the series has been in continual development for 5 years already and is still far from being finished.
And so the decks continue…
If you made it this far, thanks for reading till the end! I hope it shows that on Jalup, a flash card is not just a flash card. That I take things very seriously and pour an unimaginable amount of blood, sweat and tears into what I do here. That maybe this is a tool worthy of your Japanese journey.
Or not. They aren’t for everybody, and that’s completely fine. But for those that do use the decks, I hope you enjoy them. It’s an insane amount of work but luckily I love doing it and seeing the positive effect it has on people.
Latest posts by Adam (see all)
- Jalup App – Kanji Assist - 08/20/2019
- Why Studying with an SRS App is Overpowered #2: Randomness Breeds Creativity - 08/13/2019
- Jalup to 10k? – Interest Survey - 08/10/2019