Sharing my Anki Add-On that lets you review multiple words at once by generating sentences, "BeyondVocab"

@anonymoose It's just Chinese for the time being, although if there's traction I would love to add support for other languages. The key language-specific challenges have to do with segmenting sentences into words, and then the stemming and/or lemmatization of the words. Thanks for the feedback!

@markhavemann Thank you for the very thoughtful response. You raise some great points that learners should be aware of, and I agree with many of your premises, although I think I didn't sufficiently convey some of the flexibility (and intended use pattern) of the add-on, BeyondVocab (just going to call it BV below). I appreciate the opportunity to lay out these thoughts in some detail!

First, as relates to isolation and context:

• I totally agree that a central feature of Anki is isolation, and that BV forcibly undermines isolation by arranging your study words into sentences (or by highlighting them in your defined text, see below). Actually, we could even further say that context is an *enemy* in Anki, as the existence of its "leech" card status reveals!
• Language comprehension is inherently contextual, though, and nobody would advocate for a curriculum devoid of reading and listening practice. All of us who have successfully used flashcards also see the place for isolation practice. So, I should have been more clear that I see BV as complementing, not replacing, Anki's isolated flashcard reviews, as you suggest. (I actually recorded a study session yesterday that starts with normal card review, and ends with reading in BV)

Getting to what I see as the core of your criticism: why smash an opinionated context-based method (my BV add-on) into an opinionated isolation-based method?

• The advantages I see are primarily (1) that Anki has an established, multi-platform back-end with a thriving deck and add-on ecosystem, which all makes an add-on more convenient from a development and usage perspective compared to something standalone, and (2) that, well, reading comprehension naturally breaks down into word comprehension and vice-versa that word-comprehension naturally builds up to reading comprehension. In a sense, BV offers a "translation layer".
• The disadvantages I see are mostly (1) the potential for beginner user confusion as you point out and (2) that, since recognizing words in context is easier than in isolation, a certain inappropriateness to fully credit someone in Anki for a vocabulary review when they've read a sentence containing the vocabulary. (In the end this is a trade-off that's up to the user in BV: use the same deck for both isolated and contextual practice, or keep separate decks.)

Ancillary to the above, you ask an interesting question that mixes some concepts, writing: "Why review 100 words in 10 minutes in Anki when you can review 10 words in 100 minutes, all while reading disembodied pieces of text, which are not necessarily relevant or interesting to you[?]" The first question about efficiency is worth getting into. The second claim about the text personalization (I'll just call it), I think, is from miscommunication on my part.

W.r.t efficiency:

• With Anki, I can review 100 cards in 11.5 minutes. (My average reviews (Chinese word on front, English meaning on back) is 6.9 seconds/card (this stat from 280k reviews and 530 hours in my current, main, 5900 card Chinese deck.)
• With BV, I can review 100 cards in 8.3 minutes. (This is using a ~5 seconds/card average that I have noticed over the past few weeks of use.)
• The BV Add-On seems more efficient... except that it's a little like comparing apples to oranges. In Anki, the majority of those cards are due for review, with a high average level of difficulty. In BV, you read passages that mix known and unknown vocabulary, and the known vocabulary decomposes into cards due and not due for review, in total with an unknown average level of difficulty.
• I would guess that BV is less efficient than Anki when judged on a "due cards per unit time" metric, which I believe is your claim, so I agree. But there is an obvious trade-off that Anki is not giving you reading practice, either.

W.r.t personalization:

• I certainly should have highlighted the option for people to copy in large blocks of text (book chapters, news articles, etc.) to read inside the platform. I didn't because I'm not framing BV as a reader app; there are many good ones out there. Mine may have better Anki integration, but it certainly also has worse media integration.
• The AI-generated reviews certainly can be be a succession of disembodied sentences, if you choose, but it can also be a hyper-personalized narrative. Not only is it seeded with your Anki words (that presumably you find motivating to learn), but you could go so far as to paste in your resume and bio and ask for a story about yourself. (I have done this; it is engaging.)
• The potential from the AI-generated flow is exciting to me, since it is highly flexible. You could also, for example, copy in a news article you'd love to digest, but ask the AI to write you an HSK-3, 1000-word summary of it.

Again, thank you for the thoughtful feedback. Judging from my own experience, self-motivated language learners end up with strongly-held learning philosophies, which is good for the community because it leads to a diversity of approaches. I agree that Anki can be very efficient inside a session, but it's not a substitute for reading practice. I'm hoping the BeyondVocab add-on makes it easy for people to start supplementing and complementing their flashcard review with engaging and on-level reading practice while getting credit for and keeping track of what they've done. The key, as you said, is to complement, not to replace!

Hey everyone, it's been almost a year since I posted this thread, and I wanted to share some updates.

First, rebranding: BeyondVocab is now called Pindu (from 品读, "to read carefully and with appreciation" as I would translate it). There's a new website www.readpindu.com, and the free add-on is at the same place on AnkiWeb.

I've been developing actively most of year, and the add-on has progressed a lot in terms of polish and functionality (continued thanks to @markhavemann for feedback). Some highlights:

• Much better TTS for reading passages and words.
• Support for multiple AI providers, no longer locked to OpenAI.
• Better internal dictionary for lookups, including "meaning in context", and "chat about this word" features.
• Reader customization (fonts, colors, line spacing, light/dark mode, markup) with savable presets.
• More stable after dozens of bug fixes.

The core idea is still the same --- read your Anki flashcards together in natural language passages instead of reviewing them individually --- but the execution has come a long way.

I'm still learning Chinese and still super excited about this tool, so feedback from real users means a lot. Happy to answer any questions here.