A deck of neon flashcards. Anki cards might not be quite as fancy!

From ChatGPT to Anki : Instant Potted Vocab Decks!

With cutting edge AI galvanising the language learning world, traditional tools like Anki – which would have been considered the leading edge not that long ago – seem well in the shade. But it’s not a question of either-or. Traditional and new tech can work in happy symbiosis to support language learning.

Preparing for a recent high-stakes language mission (OK, island-hopping hol!) to Greece, I wanted to turboboost my Greek vocab. Anki was my tool of choice, of course, but one question remained: where to source new flashcard decks? Large Language Models like ChatGPT and Bing were easy choices for generating topical vocab lists, but how much copy-pasting would that involve? I wasn’t keen on spending hours formatting cards manually.

Thankfully, ChatGPT Plus’ Advanced Data Analysis mode can provide a bridge between old and new. Forget that slightly intimidating title – the main boon is simply that this mode can output a text file. And, given the right format, Anki can take such a text file as an import source. With a bit of prompting prowess, we can automate the whole process – from topic to cards, in one fell swoop. Before long, I had a fresh daily drip-drip of new words and phrases, a real shot in the arm for my Greek pre-trip.

Here’s how to task ChatGPT with the whole job of Anki deck creation. If you don’t have the Plus version, no problem – scroll down for a modified version that works with completely free plans and services!

Automatic Anki Decks – Plus Style

First of all, start a new chat in ChatGPT, and make sure Advanced Data Analysis is selected in the drop-down menu under ChatGPT-4.

Selecting Advanced Data Analysis mode in ChatGPT-4.

Selecting Advanced Data Analysis mode in ChatGPT-4.

Now, we’re ready for our prompt. Like our AI speaking prep worksheets, the beauty of this is just how specific you can make your flashcards. The topic can be as broad or narrow as you like. Here’s a sample prompt to create a French deck on the talking point ‘social issues’:

Hello! I’m learning French, and I’d like you to create an Anki flashcard deck to help me. To import a deck, Anki requires a CSV file format with just a “Front”, “Back” and “Tags” field corresponding to the English. the target language phrase and the part of speech. There is no need for header fields, so the first line should represent the first vocabulary item.
Can you create such an Anki-ready list of 50 flashcard items on the topic “Social Issues” for me, then save it for me as a .txt file I can import into the app?
– Provide a good mixture of essential and useful nouns, verbs, adjectives, and useful phrases / sentence frames (ie., so it’s not just a list of nouns!).
– Provide each term in its dictionary form if appropriate, indicating gender, plural and essential or irregular parts briefly as per convention where applicable.
– Ensure that all terms relate to the contemporary culture of the target language country as much as possible.
– Please draw on resources in the original target language when researching which words will be most useful, cross-referencing with all available data and checking constantly to make sure that the target language for the flashcards is accurate and colloquial, never bookish or unnatural.

Limitations (For Now)

One limitation with the Advanced Data Analysis mode is that it can’t run concurrently with ChatGPT’s now restored web-connected mode, or Browse with Bing. All that means is that it will be relying on its banks of training data for the vocab collation, rather than the web. But in most cases, it shouldn’t make too much difference given the vastness of that data (although it will notify you apologetically about it – see below). We’re waiting for the day – hopefully soon – that OpenAI allows users to run several premium features together.

ChatGPT Plus whirring away creating an Anki deck.

ChatGPT Plus whirring away at an Anki deck. Quirky repartee not as standard, but provided by special request thanks to custom instructions! I like my AI cheeky.

Into Anki We Go

One you have your ChatGPT-infused vocab file ready, you can import it straight into Anki. In the Anki desktop app, head to File > Import, and select the file you saved. The import settings window will pop up, including, crucially, which field matches to which column of your data under Field Mapping. The app guesses correctly for the most part, but occasionally you may need to specify that the third column (part of speech) maps to the tags field.

Importing CSV data into Anki decks.

Importing CSV data into Anki decks.

And that’s it. You should get a brief report of the number of items added, and they’re ready to play with straight away. Instant, fresh vocab decks in seconds!

No ChatGPT Plus? No problem!

Now, the above is all very well if you have ChatGPT Plus. Many platforms lack the file output side of things. But you can still get them do the heavy work of vocab-hunting and file-formatting; all you need to do is the final copy-paste-save.

Here’s how to alter the prompt for plain old vanilla ChatGPT and Bing, coaxing it to provide Anki-ready output. I’ve also made the format a little clearer, which might help if you’re using slightly older models like ChatGPT-3.5.

Hello! I’m learning French, and I’d like you to create an Anki flashcard deck to help me. To import a deck, Anki requires a CSV file format with just a “Front”, “Back” and “Tags” field corresponding to the English. the target language phrase and the part of speech.
Can you create such an Anki-ready list of 25 flashcard items on the topic “Driving a Car” for me? Output the CSV data as formatted as code so I can easily copy-paste into a text file for Anki.
– Don’t include header fields in the CSV – the first line of your output should be the first vocabulary item (ie., car,la voiture,noun).
– Provide a good mixture of essential and useful nouns, verbs, adjectives, and useful phrases / sentence frames (ie., so it’s not just a list of nouns!).
– Provide each term in its dictionary form if appropriate, indicating gender, plural and essential or irregular parts briefly as per convention where applicable.
– Ensure that all terms relate to the contemporary culture of the target language country as much as possible.
– Please draw on resources in the original target language when researching which words will be most useful, cross-referencing with all available data and checking constantly to make sure that the target language for the flashcards is accurate and colloquial, never bookish or unnatural.

Your platform should spool out some easily copiable code. Simply paste this into a text file, save, and import into Anki.

Even using 3.5, I got some great results featuring practical, useful vocabulary sets.

Creating Anki decks with the free ChatGPT3.5 model.

Creating Anki decks with the free ChatGPT3.5 model.

Experiment, Experiment, Experiment!

As with all AI prompts, it’s worth experimenting with everything to tweak, improve and get the absolute best out of it. The number of cards, the mix of words and phrases, the source of the material – make it your own. When you have it just right, you can create cards for your own, or your students’ learning, in seconds.

Oh, and don’t forget to save your perfect prompts somewhere you can copy-paste them from later, too!

If you’re keen for more artificial intelligence tips to boost your learning, please check out my book AI for Language Learners. It’s packed with practical examples to fuel your linguistic adventures!

Anki Image Occlusion with an IPA chart

Anki Image Occlusion : Lessons from the Medics

It was a long time coming, but I finally did it: I started using Anki for something other than foreign language vocabulary. Anki is steadily creeping into the rest of my life.

I know. What is the world coming to?

Admittedly, the new subject wasn’t a huge leap. I’ve started creating flashcards to drill terms and concepts from linguistics. That said, it does represent quite a departure from the way I usually create drill lists in the app.

The chief difference is the complexity of each chunk of learning material. Rather than one-to-one word and phrase combinations, we have terms with much more complex, interlinking definitions. And however brief, the information is a lot trickier to condense than simple vocabulary. Some of my flashcards were looking decidedly clunky.

If only there were some way to make it all a bit more concise and economical.

Anki-nspiration

So where to look for flashcard inspiration? Well, as it happens, language learners haven’t completely monopolised the Anki world. In fact, the app has quite the double life as a tool for medical students learning, amongst other things, terms and complex definitions!

It certainly pays to see how a diverse bunch of people use the same tool. We can learn a lot from users in other fields. And, nestling amongst the sprawling web of Reddits, there is a ton of general advice on optimising your cards.

Perhaps the cleverest trick of medical Anki users is the use of imagery for testing. Now I’m not talking about simple, one-to-one picture-word correspondences. Ohhhhh no. Medical students take it to another level, condensing lots of information into a single tableau. But to do that, they need to enlist some extra help.

Image Occlusion for Anki

The Image Occlusion Add-On for Anki allows for some quite sophisticated multi-field labelling questions. Obviously, these are ideal for drilling parts of the brain or major arterial pathways. But they lend themselves to pretty much any topic. If you can cover it up, you can turn it into an image occlusion activity.

For instance, you might think that linguistics is a rather text-heavy subject. Difficult to find too many diagrams to label, perhaps. But with a bit of creativity, you can adapt anything to fit the mould. Here’s an image occlusion activity I put together to drill the IPA consonants table and manner / place of articulation features:

Anki Image Occlusion with an IPA chart

Anki Image Occlusion with an IPA chart

Table-based data is actually perfect for these kinds of activity. And if you remember things quite visually, as I do, then making image activities out of them can yield some great memory results.

Fortunately, the Image Occlusion Add-On creators have provided a raft of training videos to learn how to use this incredibly useful tool. And – I’m very relieved to say – it’s not particularly difficult to get to grips with at all.

It is easy to forget that the Anki universe is quite massive. There is a huge amount of inspiration out there beyond our little bubble. Thanks, medical students, for pointing out this particular path!

Cross-referencing vocabulary in Excel after some tidying is applied with IFERROR.

Excel for Polyglots: Comparative audits to keep languages in sync

Duolingo, Memrise, Anki, Microsoft Excel. Huh, wait – Excel? How is that a language learning app?

Well, the Office software has some handy features that just happen to be right up our street as language learners. Namely, the ability to curate and administer lists in table form. And it just happens that this can be particularly useful if you learn more than one language.

One source of frustration as a polyglot learner is the discrepancy of vocabulary level between languages. This can be most obvious with fairly close language pairs. For instance, when practising Icelandic, I often realise that I know a term in Norwegian – but not the language I am trying to speak.

So how best to address these discrepancies?

Language auditing

Getting into the habit of performing a regular language audit, such a revisiting beginner materials is a good strategy for any learner. But one particularly powerful method for multi-language learners is the comparative audit.

In short, a comparative audit is simply taking stock of which words you know in one language, but not the other.

At the very early stages of learning a language, this can be as easy as scanning down a list. But when you get to the point of having hundreds and hundreds of words in your vocab store, the task is mammoth.

Enter Excel, data wizard!

Microsoft Excel and VLOOKUP

Most of us will have used Excel or another spreadsheet program at some point. But like me, you might not have gone beyond basic numerical information and a few simple sum functions.

It turns out that Excel is pretty good at handling textual data too. Are you thinking what I’m thinking? Yes, vocabulary lists! And it has a special function, VLOOKUP, which allows you to compare data between two tables. Sounds just perfect for our comparative audit.

Here’s how to enlist Excel to your polyglot cause in a few simple(-ish) steps.

Step 1: Port your data into Excel

First things first – you have to get your vocabulary data into Excel. The easiest way is to export from your program of choice as a CSV (comma-separated values) or tab-delimited text file. If you use Anki, this is as easy as heading to File > Export and selecting ‘Notes in Plain Text (*.txt).

Ensure that you only export the basic data and no media or tags. Ideally, you should just be exporting a word and definition / translation field. My Norwegian and Icelandic decks, for example, are populated by vocab notes with an English and Target Language field.

Export a separate file for each of the two languages you want to compare. In my case, I end up with two files, norwegian.txt and icelandic.txt.

Exporting data from Anki

Exporting data from Anki

Step 2: Import your vocab into Excel

In Microsoft Excel, create a fresh spreadsheet document, and head to File > Import. Select Text File, hit Import and locate your first exported vocabulary file from above. To preserve accented characters in our Anki list, select Unicode (UTF-8) as the File origin.

Importing vocabulary into Excel

Importing vocabulary into Excel – note that ‘Unicode (UTF-8)’ has been selected as the file origin to make sure accented characters are handled correctly.

Create a second sheet in the same document, and import your other list of vocabulary into that. You should now have a two-sheet spreadsheet document, each sheet showing a list of words in a different language. For clarity, make sure you name your sheets too. Simply double-click on the tab titles “Sheet 1” etc. to do that.

Step 3: Format your lists as tables

In each sheet, click and drag across the table to select your whole vocabulary list as a block. Now, click Format as Table in the Home section of the function ribbon / toolbar. It doesn’t really matter which style you use – I choose the colour I like best!

Once that’s done, change the new column headers to something more meaningful than the default values. I use English and Norwegian in my example below. One caveat – you need to have a column with the same title in both your tables for the VLOOKUP trick to work. Here, English will be my common column between Norwegian and Icelandic.

Vocabulary data formatted as a table in Microsoft Excel

My Norwegian vocabulary data formatted as a table in Microsoft Excel

Now, instantly, these is already more useful to us than static lists. Formatting as a table means you can use the column heading drop-downs to sort and filter your entries. Try it – sort alphabetically on the target language column. You’ve turned your data into a nifty dictionary! Not our primary goal, but a nice trick on the way.

Before we go on, it’s a good idea to name our tables so they are easy to refer to later. To do this, click anywhere in your table, then switch to the Table tab in the ribbon / toolbar. The simpler, the better – below, I just call mine Icelandic.

Naming a table in Excel

Naming a table in Excel

But now it’s the turn of our star, VLOOKUP. This is where the real magic happens.

Step 4: Adding a comparative column

Click on the target language column header of your second table and copy it (CTRL + C). Now, go to your first table, select the cell next to the target language column header (C1 in my example), and paste (CTRL + V). It should add a blank new column within that table. Let’s fill it up!

In the first cell under that new column header, we type in our VLOOKUP formula. This will depend on what you have named your tables and sheets, but mine looks like this:

=VLOOKUP([@English], Icelandic, 2, 0)

Let’s dissect that just now. The first item in the brackets is the column of the first table we’ll use at the lookup – the English entry. The second item, Icelandic, is the table we’ll look for a value in. Remember, we named that table a little earlier. The third item, 2, is the column number we’ll look for that item in, which is the target language column of the Icelandic table. Finally the fourth value, 0, is a flag to Excel that we want exact matches only.

If that boggles, simply start typing =VLOOKUP( in the cell. That calls up Excel’s formula hints and point-and-click formula building, which should help you tie things together accurately.

After doing that, something special happens – suddenly, the whole column is filled with entries. If the English term was found in the Icelandic table, the corresponding Icelandic word is pulled in. If not, we simply get #N/A.

A quick note if that doesn’t work immediately: check that the data type of the cells in that third column are set to format as General, not Text.

A cross-referencing table in Excel using VLOOKUP

Our first step in creating a cross-referencing table in Excel using VLOOKUP.

Not very tidy, is it? That #N/A is simply stating that the lookup resulted in nothing at all.

Step 5: Tying off the loose ends

We can make it all look better by wrapping it in another Excel formula, IFERROR. Change the formula in that first cell to:

=IFERROR(VLOOKUP([@English], Icelandic, 2, 0), "-")

This tells Excel to carry out our VLOOKUP function, but to return a dash if it results in an error (i.e., no data). Suddenly, it’s looking a lot neater.

Cross-referencing vocabulary in Excel after some tidying is applied with IFERROR.

Cross-referencing vocabulary in Excel after some tidying is applied with IFERROR.

Now it is crystal clear where you know a word in one language but not the other. To make things even clearer, click the dropdown on that third column, and filter it to show just the dashed elements. There is your list of words to work on in the second language!

Filtering your vocabulary items in Excel.

Filtering your vocabulary items in Excel.

Alternatively, filter on everything but the dashes to revel in the wealth of words you know in both. Enjoy that moment of pride!

For reference, here’s an example Excel file comparing sample vocabulary in French and Spanish.

Where to go from here?

What you do next is up to you. But now, you have the data in your hands, and data is power: what you know, you can act on. Export the filtered list of gaps to work on learning missing vocabulary in any number of ways.

Clearly, you can take these techniques a lot further, too. Currently, the table only checks one way, such as Icelandic to Norwegian in my example. But you can experiment with the same techniques to create much more complex and comprehensive spreadsheets to interrogate both ways.

Lastly, I’ve used Microsoft Excel in this example, but the same functionality is available in other spreadsheet programs, too. The free alternative Google Sheets, for example, has its own VLOOKUP function that works in an almost identical manner. Play around with the tools available, and you can add that dull old spreadsheet package to your list of exciting, innovative language apps!

Have you given this trick a spin? Have any interesting and useful variations on it? Please share in the comments!