I'd like to be able to click unfamiliar words in Chrome and automatically create notes in Anki 2 using an online dictionary. It'd also be nice to have an automatic method for sending text and images to Anki notes straight from Chrome. For example, if I read an article here that I want to remember, I'd be able to highlight the title, send it to Anki, and when I review, I'd see the title on the card's front with the reverse being a link to the source if I forgot what the post was about.

 

I found some Chrome extensions that purport to do this sort of thing, but didn't get any of them to work with Anki 2. Is anyone currently doing this, and if so, what is the solution?

New to LessWrong?

New Comment
32 comments, sorted by Click to highlight new comments since: Today at 12:38 PM

More broadly, I'm interested in hearing about the workflow of those who use Anki or some alternative regularly. I've used it intermittently but never felt like I was using it very efficiently. It may be that making cards always feels like that and SR's efficiency makes up for it but I'm curious to see how people have systemized the process if at all.

Also, this probably belongs in the open thread.

I use Anki to learn material from textbooks, with great results. Creating the cards takes time, but so does summarizing the material (as Luke suggests); and the benefits associated to a SRS (specifically, the testing and spacing effects) make this approach clearly superior to any of the alternatives I tried.

Briefly, I write one note every paragraph, with a question on one side whose answer, written on the other side, is the main idea of that paragraph. Sometimes a note summarizes the content of more than one paragraph, if discussion of a single idea is spread over multiple paragraphs. Some textbooks include chapter or section summaries, and when they are sufficiently detailed, I'll rely on these to create the cards instead. Here's a sample card from Peter Gray's Psychology:

Front: How are ultimate explanations of behavior different from, but complementary to, proximate explanations?

Back: Ultimate explanations are functional explanations at the evolutionary level. Proximate explanations are explanations that deal not with function but with mechanism; they are statements of the immediate conditions, both inside and outside the animal, that bring on the behavior.

(This textbook is a particularly good example of "knowledge ready for ankification" because each page comes with 1-3 marginal "focus questions" intended to guide the student in the process of learning the material. So in this particular instance one can simply write these questions on the front side of each card, as I did.)

I also create a separate card for each term or expression that is explicitly defined and whose definition I don't already know. Many textbooks include a glossary, which simplifies the process of writing down these definitions. Here's a sample card, from Ivo Welch's Corporate Finance:

Front: What is annual percentage yield (APY)?

Back: APY is the simple rate of return, often simply called 'interest rate'. Banks sometimes use the expressions 'annual equivalent rate' (AER) and 'effective annual rate'.

Is your Psychology deck public? I'd really like to have a look at it if so.

It's not public but I'd be happy to make it so. Unfortunately, I spent more than five minutes trying to find a way to share decks on Anki 2 (the version I currently use) without success. According to the Anki 2 user manual, "To share decks with other people, synchronize them with AnkiWeb, then click on "Share" from the menu next to the deck you wish to share." But although my deck is synchronized with AnkiWeb, I can't find the 'share' option or the corresponding menu. Do you (or anyone else reading this) have experience sharing decks with Anki 2?

It's in a drop-down menu default to "Options" on the right side of all my decks in the "Decks" page. Note that you need to go to AnkiWeb via https (not http) to get the version 2 stuff. (It will have a notice on the bottom about using Anki 2 if you're using the right one.)

Cool! Thank you.

There is now an entire post listing all my shared decks, as well as decks by other LW users.

Good idea. I'll add my own small contributions.

Not sure if this is what you meant by "systemized", but here's my basic workflow for textbooks:

  1. Read chapter (or, more likely, some 4-5 sections of a chapter)
  2. Summarize/analyze chapter in an emacs org-mode document
  3. Generate anki cards from the summary
  4. (Optional) Expand summary with notes from lecture

Writing new cards takes a long time, so I try to spread out the work. Roughly speaking, one textbook chapter usually takes two days, and will generate 20-30 cards. (This is for physics and math, where I'm generating a lot of cards that are relatively basic formulas and constants.) Also, most of the work is in reading, understanding, and summarizing. Making the Anki cards does take time, but it tends to be less than the other parts.

I haven't done much to automate the process, although I'm working on autogenerating cards from the emacs documentation to learn emacs shortcuts. The most important modification I made to vanilla Anki was writing a very kluge-y plugin to allow full use of my commonly used emacs keybindings (mostly movement, killing, yanking, and deleting).

A few quick pieces of advice:

  • Learn the shortcuts. It's a lot less painful when editing to type Cmd-T + M to get into Latex math mode than it is to use the mouse.

  • On a related note, if you're doing math, physics, or anything else with formulas, learn Latex if you don't know it. It's for more pleasant to review cards with pretty formatting than with ugly formatting. It also makes cloze deletion of formulas a lot easier (although I'm not sure how effective Cloze formulas are yet).

  • Batch the steps of whatever process you choose. So, I do all the reading, then all the summarizing, then all the anki additions, then all the anki reviewing. It's much faster, and way less painful to review.

  • I've found it to be easier to review old material and learn new material at the same time. Learning 10 or 20 new cards can sometimes be frustrating, especially when I basically wanted to brute-force memorize (as I did with trig formulas before starting Calculus, where I didn't care how they were derived). Reviewing, on the other hand, is generally pleasant, since I get to feel accomplished and intelligent.

  • Think hard about what information is useful. I wanted to memorize the derivation of the formula for a ring of charge in physics based on the formula for a point charge. I initially tried to make a bunch of different cards that sequentially went through the steps of the derivation. What worked, though, was having one card for the general strategy and another card for the bounds of integration. Once I had those two pieces of information, I was easily able to reconstruct the whole derivation. Since there were 3 or 4 additional similar derivations, recognizing those pieces of information as being critical saved something like 15 or 20 cards, which is huge.

If you haven't read these yet, here's a list of 20 rules of formulating knowledge, with an emphasis on SRS. This is a much longer article by the same guy, covering basically the same information in more detail. I agree with gwern, though, that Cloze deletions are not the most effective way to learn.

[-][anonymous]11y20

emacs shortcuts

I haven't started learning emacs yet but this is in my bookmarks and may be of interest to you: interactive shortcut tutorial

Thanks for that.

That twenty rules article (in particular Cloze deletions) seem quite useful and have good examples! Thanks for the recommendation. What would you suggest beyond Cloze deletions?

I don't have a definite alternative to Cloze deletions yet. I think that they are very effective in certain contexts (the example from the article of using overlapping Cloze deletions to memorize an alphabet generally seems like a great use.) However, I've found that they aren't helpful for me when trying to memorize, say, mathematical formulas. In particular, I feel that I start to "learn the deck", rather than actually learning the underlying material, and I never get a feel for the formula as a whole. So, I should really revise my comment about Cloze deletions to be "Cloze deletions are not the most effective way to learn many subjects, including those I am trying to learn."

The nice thing about Cloze deletions is they're quick to make and quick to review. I would suggest making a few sample Cloze cards and seeing how it works out for whatever subject you're studying. If they seem effective, great! Definitely use them in that case. If not, experiment with alternatives. Sorry I can't give a better answer than that.

If you haven't read these yet, here's a list of 20 rules of formulating knowledge, with an emphasis on SRS.

You can download Alex Vermeer's "20 rules" deck and use Anki to learn how to better use Anki. ;-)

Thanks. This is exactly what I meant by "systemized."

[-][anonymous]11y30

I’m interested in discussing this as well. I recall a LW thread asking for links to SRS forums and such but can’t find it now.

I have issues with formulating cards. The idea of creating a suboptimal card that I will then remember forever is so daunting that I often give up.

For example I found cloze deletion convenient, but gwern’s review advises to use free recall instead, which I am unsure what is exactly. Or I get obsessed with automating the process, like pulling definitions from a dictionary, google translate, or pydoc. Math concepts are also hard to break down into flashcard-sized chunks. It's all so agonizingly tedious.

I sound whiny. Please inspire me with SRS success stories.

I have issues with formulating cards. The idea of creating a suboptimal card that I will then remember forever is so daunting that I often give up.

As long as the information is correct, I see no problem. The reason to optimize the wording of a card, AFAIK, is to make it easier to learn. You can always make those optimizations later, while reviewing, or when looking at the statistics to find cards that you have a hard time to remember.

[+][comment deleted]11y10

. wrong click, couldn't find delete - posted above.

[-][anonymous]11y00

Each time you find content you like, just copy-paste it in its raw form to a new Anki card and don't do further processing. Eventually, when you are studying Anki and this card shows up, you can then edit it and turn it into a properly formatted card that can be studied.

highly discourage this method. "reviewing" starts taking too long and one starts procrastinating.

i copy and paste stuff i want to learn to a text file, and when i have down time go into it and turn it into question/answer/tag.

once everything is done i import it into anki

[-][anonymous]11y00

highly discourage this method. "reviewing" starts taking too long and one starts procrastinating.

My method is better than processing each card you decide to add, which breaks your reading workflow and discourages you from adding more cards.

i copy and paste stuff i want to learn to a text file, and when i have down time go into it and turn it into question/answer/tag.

once everything is done i import it into anki

This is a good workflow. Definitely better than mine. However, I don't think it will work well for cloze cards.

I do this too, but I find I rarely have time for importing.

Edit to add: for vocabulary words I've come up with a set of scripts and vim macros that put the definition (from wiktionary) and some tags into a csv file for importing into Anki. But even with this amount of automation it's more work than I normally want to deal with.

what do you mean rarely have time for importing? going to anki import file it takes maybe 30 secs?

[-][anonymous]11y40

highlight the title, send it to Anki, and when I review, I'd see the title on the card's front with the reverse being a link to the source if I forgot what the post was about

I'm not sure how effective this would be, you may want to look into Incremental Reading instead. However, for this specific workflow you could use a macro program. I just tried it out with success on OS X.

Since I've often found myself in similar situations, I decided to start developing a spaced repetition web application, called memoread, for importing information and links straight from the browser.

Ideally there'll also be Chrome and Firefox extensions, plus an Android interface of some sort. Currently, you can either add links directly to memoread, or through a bookmarklet.

You can check it out at http://damp-wave-1655.herokuapp.com/ . I'm planning on releasing the source on GitHub soon, once I create a separate repo for deployment specifics.

Keep in mind, the app should be considered PRE-alpha, with no guarantees of any functionality whatsoever, hence it being located on some obscure heroku subdomain, not a domain of it's own.

EDIT: Also, although in most spaced repetition software you can select a difficulty level of 1-5, this is not yet available as I have not had the time to implement the changes on the UI side.

[-][anonymous]11y10

Will it be a repository of links sorted by a SR algorithm or does it offer some way of processing the information into flashcards?

I can see this working well with article summaries, e.g. in conjunction with tldr.io.

Currently it is just a repository of links sorted by a SR algorithm. However, I'll consider pinging tldr.io for summaries, thanks for the reference.

I'm wary of implementing the flashcard behavior, as it allows users to cherry pick information, and possibly exclude more important information on a page, thereby by-passing the utility of learning the material.

Personally, flashcard usage seems to reinforce some sort of reflexive response to queries, rather than encouraging one to turn fields of knowledge into well-trodden gardens, as a neuroprosthetic should. I'm not sure whether this happens to the majority of users or not, more research needed.

[-][anonymous]11y00

flashcard usage seems to reinforce some sort of reflexive response to queries

This is my fear also. I trust SR for factoids, references and vocabulary - not so much for skill (at least, not in its default form).

I've heard that the folks at tldr.io are looking to improve their summaries index page. I'd be very excited to see a hybrid of your webapp and their service. Maybe shoot them an email?

This might be very late in the game for a reply. I use it mainly for MOOC and textbook learning as well as very successfully dumping large amounts of German vocabulary into my head.

From February I have the following deck stats (usually 1/2 hour in the morning, whilst stationary cycling - one boring high-intensity body task and one boring high-intensity mental task hack): Mature: 12094
Young+Learn: 392 Unseen: 59142
Suspended: 11528

The decks take a bit of fiddling to ensure you have the correct amount of cards setup. Initially in my first two months, I had too many 'new cards' set up. This would lead to a massive escalation in required time (peaked at 90 minutes in 2 weeks). I then cut new cards, just specified 10-15 new a day, depending on the deck. This evened me out to 30 minutes a day.

One caveat, I don't skip a day, ever. When I go hiking, I pre-learn the 2-3 days that I'll be away from any technology in order to pre-empt coming home to a disheartening deck. There are tablet and phone apps available, but usually when traveling I don't want to wake up to Anki when there is a view. :)

Currently I don't really have a workflow for card addition, although in the past I have mainly used CSV files that were nicely laid out and quality checked before importing. I've also done some BeautifulSoup scraping for website data extraction and word frequencies from books. I also use DuoLingo daily (German again), and the words and useful phrases I typically just dump into org-mode for transfer later (usually once weekly) when I have time.

Initially when creating and managing your decks I would suggest making backups often as sometimes syncing makes weird things happen (mostly media related disappearances between linux desktop client and android tablet) but other than that I love this tool.

For language word lists, I have also created a script that pushes the word or phrase to Google Translate (yes, yes, terms of service fingers in ears) and downloads and saves an MP3 locally. I know AwesomeTTS does most of this, but it is nice to have the media available in countries where internet access is at best intermittent and always capped.

I hope some of this helps. Have fun.

The problem with efficiency is that your reviews will pile up & up & up if you add too many cards too fast. In fact I've started taking short breaks in which I intentionally don't enter any new cards so that I can clear out the backlog. Even if you spend 1 whole minute on each card, that is only 20 minutes per day if you're following the recommendations & only doing 20 new cards per day.