My collection is made of 9k vocabulary notes in different languages, mostly Russian and Tibetan.
Description of the front side, from top to bottom: the translation of the word I’m trying to memorize (read aloud thanks to TTS technologies) | a picture | the sentence where I encountered the word, which is hidden.
Description of the back side: the word I wanted to recall in the target language (read aloud as well) | the full sentence.
Reasons why I deem images important (although it’s not the point of the topic):
They endow each note with a unique feeling of familiarity, and anchors the remembrance much more deeply than purely textual notes.
I often study while biking thanks to the auto-advance feature. When you have less than a split-second to look at your screen while focusing on the road, the efficiency of images is a huge boon.
These are the sources for my illustrations:
Midjourney, with the cheapest subscription. After a few hours spent tinkering with the parameters and plenty of fancy keywords, I more or less manage to know what to expect of this very powerful tool.
Artstor. Absolutely wonderful, my main source of pictures. Access granted by my university.
Google Images to simultaneously browse plenty of image banks, Wikimedia Commons, etc.
Previous pictures in my hoard.
Before embarking on the quest for the perfect illustration of an abstract concept, I try to think of a mythological, social or religious figure that could embody it. E.g. shrewdness → Ulysses, waggishness → a jester, asceticism → a fasting Buddha. For a Tibetan verb meaning “going to meet”, I pasted just that in Artstor and found a beautiful illumination depicting the Three Wise Men meeting at a crossroad. Objects and things, of course, are much easier.
It takes me one or two minutes to find, download and incorporate a picture to my note. I’m rather happy with the setup, the whole process often feels swift and entertaining. What are your tricks and tips to make this image harvesting effort more efficient? I’d give a pound or two of flesh for a nicely curated gallery of artworks tagged with keywords that would cover every single notion in all languages!
Wow, thanks for your detailed answer! The issue with all the fascinating automatic solutions you suggest is that Google Image won’t always find a good illustration for a notion, anything immaterial or abstract (an action, a noise, a smell). The result will be better if I think of something concrete that could illustrate said notion. But all these add-ons are very inspiring, I’ll take a close look at them. I guess they work best with words such as “kitten”, “apple”, etc. What I’d be looking for is an image collection where a search for “avoid”, “avoiding”, “steering clear of” yields paintings of the biblical scene “Noli me tangere”, for instance. That may simply not exist for the moment.
IA (GPT-4 for instance) is able to describe pictures very accurately, so one day what I envision will come true: all it would take is the willingness and funds to run it on the millions of images housed on Artstor or Google Arts & Culture and similar galleries (I find it more pleasant to use artworks than random pictures for my notes).
Disclaimer: Take the following with a grain of salt, it is my opinion only, and I probably have no clue what I am talking about.
Finding a concrete image of an abstract notion is a human effort and very personal effort. I think there are studies saying the picture does not even necessarily have to be closely related to help your memory. I do not know what your cards look like, but in my experience I tried to hard to find pictures for abstract concepts, and they worked to an extent, but once they graduated to a decent interval, they became a poor cue for the word I was testing for, unless I did one thing that we language learners are supposed to be allergic to. That is, add a word in my native language to the front of the card. I know this is sacrilege for most people. But I found that otherwise it just becomes leech bait if your goal is to increase the card interval over time. That whole “one picture is worth a thousand words” works against you. This happens even sometimes with fairly concrete words. I will give you an example. Unfair question, but what word do you think I used this as a picture for?
Well, I know, because I made the card. It is not love, kiss, baby, or smile. It was an Urdu word for parents (ماں باپ). It made sense when I made it, but since I was allergic to English on the front side of my card, once it appeared after a 4.7 month interval, I had no idea what it was supposed to be, even though I’m the one who made the damn card! It doesn’t show a lapse, so what I probably did was bury it until the next day, which actually works pretty well, but I went ahead and put “parents” in English on the front of the card, and I think that is a good thing, because it doesn’t become due again until 2027. Actually I can probably just delete the card at this point, it is easier to remember the word than the picture association. This is just one example.
So I guess the moral of the story is (for me at least, everyone’s cards and memory are different)
If it is hard to find a picture for a particular word, it is kind of a red flag that it will be hard to recall once you have a decent interval and that you won’t really get that time back that you spent searching for the perfect image.
Don’t be allergic to putting native language hints on the front of the card, despite what everyone says. I have grown to really like cloze cards with frequent English hints. Especially with the recent feature of nested clozes, you can do a lot of word chunking and then also put the whole sentence in a cloze! Without the hints, there a million words that can go in that blank, even with a picture. That way you are seeing the word in the context of a sentence, and they are quick to make as well, and they don’t always need a picture. And I think they work especially well for words that are abstract notions.
More related to your actual question, I tried some of the addons above (with the exception of the WebP one, looks like one I need, thanks @NameLessGO !), I found most of them a bit cumbersome to customize or no longer working, so what I actually use is a customized version of the legacy Fluent Forever multisearch, which includes a Google image search, and then copy and paste the image from the thumbnail view in the web browser into the note field in Anki. I find these easier to customize.
I made my own variant for Urdu. For my version I added a Reset button to close the tabs it opened to make it easier to clean up for the next word. You would of course change it to support whatever sites would be more appropriate for your target languages.
Thanks for your contribution to the discussion, your views are very clear and interesting! I share them wholeheartedly.
As you can see in my initial post, the template I use relies on both a visual and a verbal cue. This way I can pick a far-fetched illustration without running the risk of forgetting what its semantic link to the target word was.
Why add a picture if an ambiguous hint is provided? I believe the combination is more efficient than either element alone. And the time spent looking for the picture is in my opinion / experience compensated by the increase in retrieval ease. Even cards I fail repeatedly despite their containing a picture and a verbal hint don’t feel tedious or frustrating.
Regarding the insertion of whole sentences, I have always been a votary of that method and I have praised it above all for a long while. But truth be said, when reviewing cards, I don’t look very often at the original context where I found the target word: a glance at the hint (= translation in my native language) and the illustration is all I need to assess whether I remember it or not. The benefit of pasting the sentence where the word was found is that, should doubts arise regarding its exact meaning, it’s easier to dispel them.
The tool you created seems very useful, I might end up applying it to my needs, thanks!
