Is "Anki" a necessity?

Open Forum
21

I looked at the ESLPOD site (or maybe it’s one of many sites) and I could see they have conversations about various topics. I think the aformentioned chatGPT can definitely help with this. I can ask it to give me dialogues for various scenarios and I can continue and steer the dialogue in anyway I want. Kind of like those “Choose your own adventure” books when I was young, except now you have unlimited choices.

Or I can simply converse with it. Won’t help with the listening aspect, unless I was to have TTS read it, which would be helpful, but for listening I think one needs the authentic conversations…like in Easy German/Spanish/etc. channel. I’ve found those very helpful for natural speech to listen to.

22

Hi Peter,
I have been struggling with a very similar problem in Chinese. Here is a technique I use to improve my ability to understand the vernacular language:
I select a podcast intended for native speakers, that I don’t understand well and have it transcribed. Previously, by enlisting the services of a commercial service (AWS transcribe), but nowadays I’m just using whisper by openAI.
The output comes in the form of a subtitle file, I can import this into LingQ, together with the audio file. This way I can use the sentence mode to listen to the audio sentence by sentence. I will repeatedly play the audio, while looking at the sentence, the translation, use the dictionary, whatever is necessary. But I will not pass to the next sentence until I am able to understand the current one without looking at the screen. This can often entail 10-20 repetitions. This exercise is rather intense, so you might want to balance it with some free-flow reading and listening. 30 minutes often feel like a good workout already.
An alternative source is YouTube, either import a video into LingQ or by utilize a browser extension like Language reactor, this allows repeating the current subtitle as well.
I have previously written a bit about this approach here: Advice Please: Slow Japanese Progress Vs German/Chinese -...

On the topic:
I don’t use SRS. Not only do I find this activity inherently boring, but also unnecessary. Even in Chinese, I have never felt my rate of learning words to be a bottleneck, my inability to recognize those words in regular speech has been way more frustrating. The podcasts I import typically contain hardly any blue words. Now that I’m a bit more intermediate and deal with more rare literary words, I find monolingual dictionaries to be very helpful, because they provide an actual explanation of the word, often with example sentences, this information appears so much more meaningful than just a list of synonyms.

23

Hi Eric,

It’s not about the “conversations” (reg. a myriad of topics) per se, but how these dialogues are “dissected” by the ESLPod team (https://secure3.eslpod.com/about-us/).

Native speakers usually resort to highly conventionalized multi-word combos (tens of thousands of them) when speaking / writing. And they know the associated nuances (i.e., the language registers, the connotations / associations, the typical social contexts, etc.).

Language learners, even advanced ones, have “enormous problems” with this.

To give you an example in German that I discussed with my American colleague at work a few weeks ago:

Before leaving for the weekend, I texted in Teams: “Wir sehen uns am Montag in alter Frische!”
Then I thought to myself: “Hm, how could I say that in English?” And I didn’t come up with a satisfactory translation.

When I consulted Deepl, it translated the German sentence literally: “See you on Monday in old freshness!” However, nobody says that in (American) English.

Thanks to Google-Fu, I came across the idiom “bright-eyed and bushy-tailed”, which I had never heard of before. Therefore, I asked my American colleague. And he said that was something people used more in the 1950s, and it referred to rabbits :slight_smile:

Anyway, what’s missing from the English translation of “in alter Frische” is this:

  1. the contradictory tension “old vs fresh”
  2. it’s an oxymoron (Oxymoron - Wikipedia)
  3. it’s used humorously, ironically or even mockingly
  4. it’s used rather by middle-aged or older people. That is, children oder teenagers probably never use this expression.
  5. It tends to presuppose a certain familiarity between people who know each other.

For some background info in German see also:
in alter Frische – Schreibung, Definition, Bedeutung, Beispiele | DWDS (btw., I like the ironic answer “Alt ja. Frisch eher nicht” = “Old yes. Fresh rather not” in this context :slight_smile: ).

This is more of an idiomatic structure, but such nuances apply - mutatis mutandis - to many other collocations that have not achieved the status of established idioms as well.

And to become fluent in an L2, you need to understand such semantic and pragmatic nuances…

“I think the aformentioned chatGPT can definitely help with this.”
Yes, it should be able to create some example sentences for some collocations.

However, can it provide / explain the semantic and pragmatic nuances that are involved here? I don’t think so.
That’s usually “native speaker territory”. So, the winning combo for the implementation of such SLA solutions is, IMO, rather a tech (AI, corpus linguistics, etc.") and native speaker hybridization :slight_smile:

PS -
I would, of course, also resort to specialized dictionaries for “collocations”, but combine them with contemporary tech solutions (AI and Co)!

PPS -
Never trust blindly a native speaker: say “no” to rabbits and “yes” to squirrels :slight_smile:

“Bright-eyed and bushy-tailed are two terms that developed independently of each other. Bright-eyed supposedly comes from the late 1500s, while bushy-tailed is said to have hailed from 1865-1870, though no direct source is clear for either of them.
They were seen together for the first time in talking about a squirrel, which did, in fact, have bright eyes and bushy tail.”

24

@PeterBormann Sorry, I wasn’t very clear.

Let’s consider the following scenarios using both an ‘artificial’ and a ‘natural’ SRS. We will use units of days with 0 referring to the day 0, the day of exposure to a new word. 7 is day 7 and would be one week later.

‘Artificial’ SRS eg. Anki
Increasing intervals by doubling: 0 1 2 4 8 16 32 64 128
Fixed intervals of 5 days: 0 5 10 15 20 25 30 35 40 45 50 55

These intervals are very regular and in certain algorithms, such as the Anki one, if you get it wrong, you will see the word more frequently.

‘Natural SRS’/reading and listening and following your interests
A ‘natural’ SRS is semi-irregular in it’s intervals, or at least has more randomness involved, but may look something like this:
High-frequency word: 0 1 3 4 5 6 8 9 12 13 14 15 16 17
Mid-frequency word: 0 4 8 11 17 29 31 36 44 48 53 60
Low-frequency word: 0 40 112 190 319

As you can see here, there are intervals between every exposure to the word. The paper you referred to talks about intervals vs no intervals. A ‘natural SRS’ has intervals.

Unfortunately, I can’t read the paper and the method, so I can’t refer to the details, but I’m very sceptical if we can generalise this “relative schedule of repeated tests had no discernible impact” to “has no discernible impact.” Maybe it didn’t have a discernible impact in this, one experiment, but it doesn’t necessarily mean it’s true in the general sense. There’s chance involved, right. I mean, if you believed it to be true, then why are you using Anki in the first place? My guess is because you believe the frequency of the intervals for low-frequency words is too low, right? Especially to learn the words at a decent speed? I agree.

There are several ways to alter the ‘follow-your-interests’ approach of the ‘natural’ SRS to increase the frequency of the intervals of words:

  1. Read and listen more. Simply increase the number of hours you spend each day reading and listening. It’s not always possible for everyone, but for some it’s an option. You can also achieve this by increasing your reading and listening/playback speed. If you do this, you can still follow your interests and learn many new words.
  2. As you mentioned, be more selective in your materials and less ad hoc. Certain words appear much more frequently in certain contexts, in certain domains, and by certain authors. For instance, economic words appear more in books on economics or financial columns of newspapers. To increase the intervals of low-frequency economic words, and hence learn them, read more economics material. On LingQ, you can see the number of yellow words in each lesson, so don’t let that get too low.
  3. Reread and/or relisten to material. If someone is supportive of an ‘increasing interval’ idea, they can follow this strategy with their rereads/relistens too.
  4. Using LingQ, skip large portions of text, where you already know the words and focus on the yellow and blue words. On the browser, you press the right arrow key to do this. Steve mentioned he was doing this on his Persian podcasts.
    With such strategies available, the question is: Why do you use an ‘artificial’ SRS? There are clear benefits to more reading and listening (such as learning collocations, as you mentioned, and exposure to the word in various contexts, not just the single one on the flashcard). By using an ‘artificial’ SRS, this comes with the opportunity cost of less reading and listening.
25

Hi, Florian!

Yes, that’s a good technique. And I use(d) something similar with all kinds of podcasts, Youtube videos or Netflix in various languages (English, French, Spanish or Portuguese). But I’m not happy with this approach because the “nuances” (mentioned above) are often missing.

For example, it took me about 15 min, just to find a satisfactory answer (Deepl, Google-Fu, discussion with my colleague) for the “in alter Frische - bright-eyed and bushy-tailed” problem.

IMO, we can do better than this nowadays when it comes to (tens of) thousands of collocations (idioms included)!

“I don’t use SRS. Not only do I find this activity inherently boring,”
Well, I find the solution LingQ-to-Anki quite handy (for testing).
However, ff I had to create (tens of) thousands of Anki cards manually myself,
I probably wouldn’t use it.

But the advantages of the LingQ-Anki-combo are:

  1. I know the context from reading / listening on LingQ
  2. LingQ provides often complete sentences
  3. I can translate from my L1 in various L2s, which is much harder and, therefore, more effective than the other way around.
  4. It’s faster than re-reading texts on LingQ.

Re the “boring” part:
I often do the same things every morning: working out, running, taking a cold shower, doing Anki drills, and using LingQ.

It’s just habit - and I’m never bored. It’s the same for brushing my teeth every day… :slight_smile:

BTW, that reading is inherently more “interesting” than flashcards is a myth in my opinion, because there are so many “bad” texts out there (from didactically graded readers and depressing Assimil and Co beginner texts to AI-based text generators - “yes, OpenAI, I’m talking about your GPT3 models, too” :slight_smile: - that I prefer short and crisp flashcard exercises.

But, of course, I love reading and listening as well, but this refers usually to more advanced stuff (history, social sciences, linguistics, great novels by professional authors and not just well-meaning didacticians). :slight_smile:

In short:

  • Reading is great (and more exciting than flashcards) when the content is not only comprehensible, but compelling.
  • Reading is not great when the content is comprehensible, but produced by second- or third-rate (human or AI) storytellers.
26

Good points, nfera!

But it’s getting late and I’m hungry :slight_smile:

I’ll answer tomorrow…

27

SRS is an auxiliary and secondary part of committing vocabulary to long-term memory by deliberately fighting the forgetting curve. Acquiring vast amounts of words more efficiently depends more on our ability to make associations by encoding them into our brains when interacting with the language.

I can memorize and retent way better when I am at the intermediate level compared to the beginner level. The key things lie in comparable skills we have developed in the course of learning a language and applying such skills consciously or subconsciously in all activities that we engage in the language. These essential skills include mastery of phonology, familiarity with the writing system, ability to sound out words, use contextual clues, follow the storyline, connect with personal experience, make cultural, social, and historical references, etc. I also find that monolingual dictionaries are extremely helpful in providing precise definitions and sample sentences in context. However, sometimes I need help understanding the sentence, even with the translation in sentence mode on lingq.com.

How does everyone deal with more complicated sentences loaded with figurative speech, metaphors, cultural and historical references, or any other idiosyncrasy in the language?

28

@llearner

I agree it’s important to consider the forgetting curve and commiting words and phrases to long-term memory. The question is, Why have you chosen to use the strategy of using an SRS to do it? It comes with the opportunity cost of less reading and listening. There are other strategies (listed below), which seek to also “deliberately fight the forgetting curve” and come with the additional benefits of more reading and listening. Namely, exposure to the word/phrase in different contexts compared to the single sentence on the flashcard, and exposure to other words and phrases in the process, further strengthening your intuitive knowledge of them.

15 minutes of Anki per day is almost 100 hours per year. It adds up. Imagine 100 more hours of reading and/or listening. Even half of that Anki time is still equivalent to several extra books in a year.

29

Hi,
whether Anki or any SRS or flash card app is a good tool for you, you have to find out for yourself. Everyone learns differently.
I like to work with a SRS-system and I use it (on and off) for about 11 years now. Personally I tried Anki once (about 11 years ago) but then I chose to use the app “Flashcards Deluxe” which fits me and my learning style more and the app served me well for all this time because I spend a lot of time in trains and busses, and I always have my phone with me. Actually I liked the app a lot as a beginner for French, because there is the possibility to use text to speech for French and I could listen to the pronunciation of each expression while learning it. That helped a lot with my own pronunciation. And there is the option to write my answer down as a drawing option. I always use that tool when I am somewhere where I can’t use pen and paper, because I think that writing the words down helps learning them faster.

The advantage of an SRS app itself is that the app reminds you when you have to revisit the vocab. But people tend to feel pressured after the app tells them they HAVE to study 345 flashcards now… and sometimes people just swipe “known” without really knowing a word, just to get rid of the heap of flashcards that they “have to” learn that day. That’s not how SRS works…

Before you use an app you should consider that it can take a lot of time to create flashcards in an app. That depends on the app and the source of your learning material.
BUT, when you already have great cards on paper… just use them the same way!
Actually I switched to an app because I had a looooot of paper flash cards “sitting around” as the result of learning three languages: more than 30.000 flash cards… and I don’t want to know how many hours (or rather days!) it took to transfer the cards to the app.

30

@nfera
“clear benefits to more reading and listening”
Definitely - esp. using “ultrareading while listening”.

  1. “Read and listen more.”
    Well, there’s a major downside to just “reading (while) / listening more and more”, esp. from a B2-C1 level upwards:

In general, the mental effort involved in simply recognizing form-meaning combinations embedded in contexts is way too low compared to actually producing and using such form-meaning combos appropriately in co- and contexts.

Therefore, it’s a common complaint among CI practioners (SergeyFM was one of them in this forum a few months ago, and I’m experiencing exactly the same thing in Br. Portuguese right now) that even after reading / listening for more than 1000-2000 h in the respective L2s, we still struggle a lot in speaking (and writing).

  1. Notabene:
    When it comes to writing in our L1s, that’s even true for native speakers!
    In other words, if native speakers aren’t professional authors or have at least an academic background, they tend to be “bad writers” (esp. if they’re are also bad readers).
    And I’m not talking about “functional analphabets” in this context, i.e., almost 10 percent of the German population:

“Low-literate adults who attended school but either did not complete their education or, for a variety of reasons, did complete it without attaining the expected level of reading skill are referred to as functional lliterates. In Germany, approximately 7.5 Million adults are considered functionally illiterate.” (highlighting by me)

  1. Using “LingQ-to-Anki” is a kind of “intermediary” solution for this problem (the “real” solution is, of course, producing / using the vocabulary in oral and written communication, esp. with native speakers, but also with more powerful chatbots à la ChatGPT, - a lot):
  • First, reading-only / (ultra-)reading while listening (URL) / listening-only based on LingQ
  • Second, testing myself in L1 → L2 translations for the context-embedded URL-vocabulary using Anki.
    That is, if the Anki drills consisted only of L2 → L1 translations, then this would be another and far too easy recognition exercise. And in that case, I might as well just resort to the reading strategies you mentioned above.
  1. Apart from that (i.e., tougher "active L1->L2 recalls instead of pure L2 recognition), there are other use cases for artificial SRS. For example:
  • Learning specific grammar structures (verb conjugations, L2 cases - see the current discussion with xxdb reg. Russian, etc.)
  • Low frequency collocations that you mentioned above.
    etc.
  1. Can artificial SRS be further improved?
    Definitely - see the “interaction” of AI, corpus linguistics, specialized dictionaries / SEs for collocations, and native speaker knowledge indicated above so that collocations are enriched with linguistic and socio-cultural background knowledge.

Why?
Remember the “bright eyed… - alte Frische”-problem mentioned above?
If the assignment for learners of German (let’s say, Asad, Eric, Bembe, etc. in this forum) was to write a short dialogue between contemporary German teenagers using options such as “putzmunter”, “quietschfidel” and “alte Frische” to describe someone who is full of energy early in the morning… which options would you use?

  • Why and why not?
  • Are there other options not mentioned in this context? (yes, there are!)
  • And what do you think the sociocultural connotations of these options are?
31

@PeterBormann

Okay. I misunderstood. I thought you were referring to using an SRS to learn new words and new collocations. In particular, commiting them to long-term memory. I think this is what the majority of SRS users use it for. This is what I was referring to with the modifications to the ‘follow-your-interests’ approach to perhaps make it more effective than an SRS.

But you are referring to using the SRS as a stepping stone to speaking/writing by only doing L1 → L2 production. You mentioned Comprehensible Input learners having the problem at B2/C1 of struggling to speak. You also mentioned yourself that the most effective solution would be speaking with and writing to native speakers or AI. This leads to the next question, Why don’t you do that to improve your speaking/writing instead of using L1 → L2 production with the SRS?

I, too, don’t believe that you can become a fluent speaker by only input. Even Stephen Krasher I think doesn’t believe this. One of Steve’s phrases that he loves to say is: “If you want to speak well, you have to speak a lot.”

32

Interesting discussion! Myself, I’ve been a convinced Anki user for a long time. Not only for language learning but for almost anything.

However, since I started here on LingQ (not long ago), I experience really quick progress in Italian. I also have been using Memrise since I started with the language, but soon after starting here, Memrise felt like a waste of time. It just feels slow, and intuitively I think it’s more effective to invest that time in immersing instead.
As for Anki, I removed the deck with the first 1500 vocabulary today as I spent daily almost an hour, going through vocabulary and short phrases, I actually know already. - Again, feels like a waste of time.

What I highly enjoy is going through mined sentences from movie or video game trailers. But creating these cards take quite long in comparison. I guess I will still keep creating them at times, I want to slow down a bit.
The slow pace is welcome though in other languages which are a bit more difficult to learn. At least for now.

About the SRS - I like that using LingQ’s built-in SRS saves a lot of time. However, I only activated multiple choice. As many words have various meanings depending on their context, I think, the other options just complicate things and I believe, with time, the needed vocabulary will eventually be learned anyway, as we keep encountering them.

Maybe if I wanted to build some vocabulary in a specific area, I would use Anki again.

33

@nfera
“Why don’t you do that to improve your speaking/writing instead of using L1 → L2 production with the SRS?”
There are two reasons:

  1. Such Anki drills are less time-consuming than speaking / writing in Br. Port. at the moment.

  2. It’s an experiment: to what extent do these Anki drills make speaking/writing easier (and more meaningful) in the L2?

Another interesting approach may be just to use ChatGPT or the combo " L1-L2 drills (based on Ultra-Reading while Listening vocabulary) plus ChatGPT".

I’m not a fan of ChatGPT “stories”, but using AI-generated dialogues based on URL-vocabulary could be an interesting option (see the current thread reg. ChatGPT).

34

@PeterBormann

For the experiment, fair enough.

Regarding the time-consumption, they may be less time-consuming, but as you mentioned, they aren’t as effective as speaking/writing to people (referring to your comment of the ‘real’ solution to the problem). I think it should be thought of as how effective something is per minute/hour invested.

That said, if you only have 15 minutes, it’s reasonably fast to log on to ChatGPT for 15 minutes and have a texting chat. It’s also relatively fast to open up Google Maps and write a review of the restaurant you just visited in your L2. Regarding tutors and what not, yeah, you need a bit bigger time block and a little organisation, generally.

Regarding your comment about “at the moment,” my experience with Italian is that my first 5 or so hours of speaking (so maybe 8 or 10 hours of conversation) was particularly challenging. It’s still not easy (I also have a limited passive vocabulary at 11k Known Words), but it became significantly easier after the first few hours. I really saw a massive jump. Maybe it might be similar for you.

35

Here’s one suggestion that’s an alternative to the SRS:

Randomly read a page in a book or graphic novel in your target language that you’ve read before.
And randomly read over a page or two in LingQ of a lesson that you’ve completed.

Some random repetition of a whole chunk of something you already have memory of can be as powerful as an SRS

36

@Peter,
Thanks Peter for the insight.

I definitely know the phrase “bright eyed and bushy tailed”. My parents would use it when I was growing up. “You need to get some sleep so you’ll be bright eyed and bushy tailed in the morning!” =)

I’m a little surprised deepl.com didn’t get “in Alter Frische” as it often has these idioms down. I wonder if given a little more context if it would’ve figured it out. Then again, as great as it is, it always has room for improvement.

I usually look in dict.cc for these idioms for German. It’s had just about any I’ve every needed to look up:

pons usually does pretty well with these too, although it didn’t have this one exactly (it referred to some contextual examples).

37

Hi Nefera,
I want to clarify one thing I do flashcards but not so much combining with the SRS technique. I use flashcards for note-taking and identify the areas I must work on to overcome a hurdle or advance my learning to the next level. Fundamental language skills and higher cognitive power to process the massive volume of information efficiently should be deliberately practiced and enhanced by consuming or producing tons of context in relevance.

I, too, don’t believe that you can become a fluent speaker by only input. Even Stephen Krasher I think doesn’t believe this. One of Steve’s phrases that he loves to say is: “If you want to speak well, you have to speak a lot.”

I couldn’t have agreed more that every core skills in the language need to be developed over time with tons of effort. A couple of things I am aware of that hinder my efficiency with the comprehension of text are pronunciation and grammatical structures. Flashcards with multiple sample sentences and authentic audio provide more concentrated and detailed information on the grammatical structure or specific knowledge about a colocation in the language. These practices are stepping stones to opening the gateway to further studying for a specific subject or at a particular language level, enabling me to consume or produce related content with much less difficulty.

I have lingqed terms tagged with “grammar,” “idiom,” “phrase,” “culture and custom,” and “metaphor,” etc. It would be a little too daunting to dive right into a book such as the one for etymology or specific colocation like Chengyu in Chinese with historical anecdotes, especially if someone’s not at that level or has no basics in the subject.

A good question is when would SRS, whether natural or artificial, become less relevant as we continue developing our core language skills with engagement in the language corresponding to the framework of Bloom’s Taxonomy?

38

Hi Florian, sorry if this is a stupid Q, but how do you use Whisper? On Introducing Whisper I cannot see any button to get it running or to uload an audio.

39

Forum threads with lots of replies are always interesting, no clue where my response might end up…

@JanFinster Re: Whisper:

The whisper repository is located here: GitHub - openai/whisper: Robust Speech Recognition via Large-Scale Weak Supervision Instructions are in the readme. But note that you essentially need a modern Nvidia GPU with lots of memory (8GB+). Since I don’t have such a system, I use a re-implementation of the original called ‘whisper.cpp’ GitHub - ggerganov/whisper.cpp: Port of OpenAI's Whisper model in C/C++ It is optimized for use on CPUs and handily outperforms the original on my system (Apple Mac mini M1). The medium model converts here in about 3x real time and the large one in about 1x.
As for installation, if you’re regularly compiling things, this should be trivially easy, if not then you might have a bit of a learning curve ahead…

A less involved way to try it out was introduced here:

The original thread that introduced whisper to the forum is here: https://www.lingq.com/en/community/forum/open-forum/best-way-to-generate-subtitles

Also, I probably should have added to my other post, that whisper is really good in terms of word error rate (WER) but pretty bad at timestamp accuracy. So, the exercise I proposed doesn’t work well with whisper’s transcripts. Commercial services typically have excellent timestamp accuracy, as do YouTube subtitles.

40

My feeling is Anki is useful at beginner level to build up a base of vocab and practice basic grammar but is better used sparingly beyond this.

It’s easy to use all your spare time just revising old words in Anki. I feel you will develop quicker if you use most of your available time reading and listening to new content rather than revising old content. This reading/listening method is a type of spaced repetition because the same words come up again and again so you keep getting refreshed.

I am lower intermediate. I spend 10-15 minutes per day on Anki. I have to be careful I don’t add too many words each day or my review time will get too long. Most of the time I don’t add any words or a max of 5 a day. Nowadays i only add words/phrases that are not intuitive or easy to remember. Basically: Anki for 10-15 minutes is a delight. Over 20 minutes gruelling! haha.