It’s pretty hacky (the script opens a bunch of google image searches so that you can download the pictures) but combined with the image occlusion anki addon, it has allowed me to memorize sets that are 3 time larger than I can normally memorize with Anki.
Graphviz must be installed and set to root, you also need the launchy ruby gem.
The program will generate a random color scheme and layout engine, which can be reassigned. Color schemes can be found here: graphviz.org/doc/info/colors.html, and layout engines can be found here: http://www.graphviz.org/cgi-bin/man?dot
3.The program will ask if you want images. If you click yes, the program will later open a bunch of browser windows equal to the amount of items in the set.
Enter the name of the graph
The program will ask for the name of the category. If you enter it, this will be the “center node”. If blank, there will be no center node.
Enter your set, one item per line. When done, enter a blank line.
If you chose images, the program will open a bunch of google image searches to find images. The images should be saved as (all lowercase version of the search with spaces removed).jpg, in the same directory as the ruby file. In order to make sure you get jpgs, you should save the thumbnail that google generates, rather than saving the actual image.
A graph will be generated.
Open the graph in the image occlusion extension in anki to start memorizing it.
One concern though: by adding colors, shapes, borders, etc., you are essentially adding extra detail/context to the memory-triggering side of the card, which will indeed improve recall when you have that detail/context available. However, in a live scenario where you actually have to remember the information, that context will likely not be available.
(An example: if you’re trying to learn the locations of US states, and you get a map where each state is brightly-colored, you should probably make the map grayscale and uniformly-saturated before you apply image clozes. Because when you actually need to know where New Jersey is, you will not be given the information that it’s red on your map.)
Then again, I can think of some hard-to-verbalize ways in which the extra detail might improve recall even when you don’t have the detail available.
Overall, I’m not sure if this is a good idea. It might be worthwhile to try memorizing (random?) sequences using these graphs for half the sequences and plain text for the other half, then testing each sequence of them outside of Anki (by running through the set mentally, say).
I actually started out with using uniform colors, shapes, etc.
I can only give my own experience, but I find that those earlier images are universally harder to remember, even when I don’t have the image in front of me and I’m just trying to recall the set on it’s own. This is true even for cards where I have only four items in the set for the uniform images, and upwards of 15 for the non-uniform ones.
I think that what happens is that these extra cues help in the initial learning and memorization. As I get better, I can simply visualize the location of the node in the image, visualize the attached image, which brings to mind the text. I have trouble getting to this point when I don’t have the other context cues to help me out initially.
I don’t quite understand what test you’re suggesting in your last paragraph. I think what you’re saying is try to memorize a random set using simply text, then a random set using simply the images, and then test myself outside of anki by trying to recall the sets. If so, I have done this, and the images (with the crazy shapes), outperform by a large margin. I can’t remember a set of more than about 5 using simply text in Anki.
Something useful to those of you who use Spaced Repetition Software:
I made a little ruby script that can turn ordered and unordered lists into easily memorable diagrams like this:
https://onedrive.live.com/redir?resid=51A281FEEAA3C35!1455&authkey=!AKtQ02Ji961f_n8&v=3&ithint=photo%2c.png https://onedrive.live.com/redir?resid=51A281FEEAA3C35!1457&authkey=!AMtC38EHOFcImTI&ithint=folder%2c https://onedrive.live.com/redir?resid=51A281FEEAA3C35!1458&authkey=!AOIm4ua5-c1TFsQ&ithint=folder%2c
It’s pretty hacky (the script opens a bunch of google image searches so that you can download the pictures) but combined with the image occlusion anki addon, it has allowed me to memorize sets that are 3 time larger than I can normally memorize with Anki.
The script requires Graphviz, as well the launchy ruby gem. It can be found here: https://onedrive.live.com/redir?resid=51A281FEEAA3C35!1459&authkey=!ACtSe9c5YnpYk9Q&ithint=file%2c.rb
Quick readme:
Graphviz must be installed and set to root, you also need the launchy ruby gem.
The program will generate a random color scheme and layout engine, which can be reassigned. Color schemes can be found here: graphviz.org/doc/info/colors.html, and layout engines can be found here: http://www.graphviz.org/cgi-bin/man?dot 3.The program will ask if you want images. If you click yes, the program will later open a bunch of browser windows equal to the amount of items in the set.
Enter the name of the graph
The program will ask for the name of the category. If you enter it, this will be the “center node”. If blank, there will be no center node.
Enter your set, one item per line. When done, enter a blank line.
If you chose images, the program will open a bunch of google image searches to find images. The images should be saved as (all lowercase version of the search with spaces removed).jpg, in the same directory as the ruby file. In order to make sure you get jpgs, you should save the thumbnail that google generates, rather than saving the actual image.
A graph will be generated.
Open the graph in the image occlusion extension in anki to start memorizing it.
Awesome, thanks!
One concern though: by adding colors, shapes, borders, etc., you are essentially adding extra detail/context to the memory-triggering side of the card, which will indeed improve recall when you have that detail/context available. However, in a live scenario where you actually have to remember the information, that context will likely not be available.
(An example: if you’re trying to learn the locations of US states, and you get a map where each state is brightly-colored, you should probably make the map grayscale and uniformly-saturated before you apply image clozes. Because when you actually need to know where New Jersey is, you will not be given the information that it’s red on your map.)
Then again, I can think of some hard-to-verbalize ways in which the extra detail might improve recall even when you don’t have the detail available.
Overall, I’m not sure if this is a good idea. It might be worthwhile to try memorizing (random?) sequences using these graphs for half the sequences and plain text for the other half, then testing each sequence of them outside of Anki (by running through the set mentally, say).
I actually started out with using uniform colors, shapes, etc.
I can only give my own experience, but I find that those earlier images are universally harder to remember, even when I don’t have the image in front of me and I’m just trying to recall the set on it’s own. This is true even for cards where I have only four items in the set for the uniform images, and upwards of 15 for the non-uniform ones.
I think that what happens is that these extra cues help in the initial learning and memorization. As I get better, I can simply visualize the location of the node in the image, visualize the attached image, which brings to mind the text. I have trouble getting to this point when I don’t have the other context cues to help me out initially.
I don’t quite understand what test you’re suggesting in your last paragraph. I think what you’re saying is try to memorize a random set using simply text, then a random set using simply the images, and then test myself outside of anki by trying to recall the sets. If so, I have done this, and the images (with the crazy shapes), outperform by a large margin. I can’t remember a set of more than about 5 using simply text in Anki.