How to learn in batches with SRS?

Chrysmc

Member
This is an amazing help. Thank you. I'm so grateful to everyone who takes the time to post. I had an awesome Skype session with Shun, which worked well for my learning style and got my basic settings in place. Now, with these settings, I can expand on them. Thanks again.
 

Chrysmc

Member
I've set up my test profiles with the settings that have been so kindly shared. Thank you Edsko and Pdwalker. You're truly wonderful. I've just run into a couple of minor issues that I'd like to ask you about.

With the Reading (New Words) profile, I don't get the option to reveal separate parts. The test settings list 'Number of choices'. The option for this is 4 or 5. And 'Choices from'. Options are All cards, Session catergories and custom categories.

In the Pinyin testing, I get a message to say 'No flashcards found'.

I'm loving the app now, especially with the variety this setup gives me. I'm not the smartest compound in the character, so any help is warmly welcomed!
 

Shun

状元
Hi Chrysmc,

too bad I can't be of much help here because I prefer the simpler "splitting categories, repeating them back and forth and setting card filters" approach to the SRS, because I feel it gives me more control.

Just a thought that came up regarding the "No flashcards found" message: Could it be that you have already studied the category using the first setting, then after switching to Pinyin testing, used the same scorefile? If so, you could try creating a dedicated scorefile for each testing profile. For that, you just switch between profiles and add a new scorefile under "Scoring" / Scorefile / Manage / New for each profile. But as I've stated, I may be thinking in the wrong direction. :)
 

Chrysmc

Member
Hi Shun,
It was your help which gave me the start to being able use Pleco at all and I'm super grateful for it! Thank you. I'm still using your concept of smaller categories and I choose multiple categories to study, depending on my current focus. I'm not sure if that's a good way of doing it or not but at this early stage, it seems to be doing the job. I was able to use the same categories for each of the other testing, even after using them previously. But I'm actually quite confused about scorefile. Can you explain it to me at all? I created the 5 different scorefile profiles (Reading, listening, writing, speaking, pinyin) but didn't enter any parameters for the them. (That sentence alone probably shows how little understanding I have!).
 

Shun

状元
Yes: A scorefile keeps scores and the history (how well you've done, or how many answers were correct/incorrect) for each card in the flashcard database. If you're good at reading, the scorefile of the "Reading" profile will have higher scores than, say, the scorefile of the "Listening" profile. Having different scorefiles lets you keep the record of your learning success separate for each discipline. This will influence the selection of cards Pleco tests you on in the different disciplines.
 

Chrysmc

Member
Ahhh, ok. So I don't need to set any parameters up because Pleco does it for me. And I can see the results in 'statistics'. My Pleco fog is starting to lift!
 

pdwalker

状元
Ahhh, ok. So I don't need to set any parameters up because Pleco does it for me. And I can see the results in 'statistics'. My Pleco fog is starting to lift!

Yeah, it took me awhile to get my head around it. This thread is what got me my clue.

After you've played with the different tests, you may change your test profiles based on what works for you. For example, when I first ran my listening tests, I would only show audio. However, I had to many problems identifying the sounds based on a single character (as many words share the same sound and intonation), so I changed my listen test to also show the characters.

Once you've gone through the settings for yourself, re-read the pleco documentation. It's a great reference, as long as you already know what you're looking at, which is true for most things unfortunately.
 

pdwalker

状元
It's been a year, and I've changed a few settings based on how I am learning.

First my learning process is to review each new card until I've gotten it right three times in a row, then use the SRS system for long term memory review and testing. Thus, for each thing I wish to test, I have to sets of test settings - one for new words and one for review.

I only recently learned that getting a card wrong resets your score to the lowest value, which is reasonable if you only use the SRS for memory training.

However, with my system, if I get a character wrong, I am forced to review it three more times correctly in a row before it goes back into the SRS rotation. Thus, I do not want to reset my scores back to a low value. Instead I'm reducing the score by 50%. My justification is, if I "forgotten" something I already know, I'll relearn it faster and remember it better the next time it comes around, so I don't want it to show up as often as characters I am in the process of learning for the first time.

Perhaps if I were starting again, I use only just use the SRS, in which case I'd use only the review test profiles with all the character filters turned off and the default scoring parameters, and forget the new words test profiles.

Notes: For the tests, I normally set the # of cards to use per session to "all". If I'm in a hurry, and I only want to do a few, I'll set a lower number and have several review sessions during the day, depending on my mood, drive, and how far behind I am. I'm I'm all caught up, I'll leave it on "all".

Since my deck has a lot of single words and characters, it can sometimes be hard to figure out what character/word I am being tested on without having context. Since I cannot give myself context, I instead also show the character and pronounciation. Take for example, 还. Is that hai2? or huan2? or xuan2? (ok, I'm cheating. That last one is really obscure), or shi4 (是? 世? 事? 士? 室?... ) I'm running into this more and more as I learn new characters/words/meanings, so for listening, and writing, I'm giving my self some extra information in order to help me figure out which character I should be recognizing, writing, speaking, whatever.

I suspect I may not be using optimal settings, so I'm always open to suggestions on how to improve my settings.

I guess this is why Mike want's to hide some of the test settings in the future. ;-)
PlecoSettings.png
 

Attachments

  • PlecoSettings.pdf
    52.9 KB · Views: 799
  • PlecoSettings.ods.zip
    18.7 KB · Views: 706
I have tried a few different learning settings, in the beginning I also used to separate new flashcards from the SRS ones.
Eventually, SRS is for me the ultimate learning tool. Whatever word I add today, I learn it within a few days with the SRS.

Say tomorrow and the day after, I'll fail the word. But the next day, I will remember it. Maybe after 10 more days I'll fail it once, but then that's it.

I figured out SRS can really take care of everything, so it reduces my learning setup.
But I know everybody is different so it's difficult to find a pattern everyone agrees on.

Using SRS only, I can learn 70 new words per week. The number has gone up to 200 a few times when I went berserk on learning. But 70 per week is the average for the last 3 years.

I think one thing you could try is keep track of the new word count (visible in the Statistics page), and compare the number of weekly characters while you try different learning methods, to figure out which one works best for you!
 

pdwalker

状元
I think I'm coming around to your way of thinking.

My original thought was to make sure I reviewed the new characters a few times first since initially I will forget them so often. If I just purely start with SRS, then forgetting a character means I won't get to review it again for another day (I think).

Hmm.. Maybe I'll turn off the card filters for SRS and just try that for a while and see how it goes.

Thanks for the feedback.
 

pdwalker

状元
Using SRS only, I can learn 70 new words per week. The number has gone up to 200 a few times when I went berserk on learning. But 70 per week is the average for the last 3 years.

That's at least 10 new words/phrases/cards per day! I'm no where near that and I struggle sometimes to even retain the memory of what I have learned.

How much time do you send doing your reviews?

Do you only use SRS just for reading retention? Or do you test yourself on anything else (writing, listening, pinyin, speaking)?
 
I only do SRS, only for reading retention.
Actually for me, doing SRS for reading helps me on all levels, except for writing (with a pen). It helps me remember the character, the way it sounds and how to type it with a keyboard.
I spend about 40-50 minutes per day doing SRS if I focus. That's about 200 to 300 flashcards per day. I think I used to be slower, but after 3 years of doing flashcards every single day, I became much faster, and my memory for learning Chinese also improved.

After I do the SRS, I fail generally around 25 to 35 cards, which I repeat until all are "correct". That part adds maybe 10 minutes to the daily practice.

Writing with a pen is the one part that I have neglected so far. In my experience (and probably due to my IT job) I never write, even English, and I feel that the time spent on writing Chinese would not be so useful in the end. Because I think it's very easy to forget the writing without practicing. The only things I need to write sometimes is my name and my address, nothing else...
 
I've been neglecting my studies for quite a while :( That said, on the positive side I lived in China for quite a while too so at least my 口语 and 听力 are much improved :) Anyway, it's time to start preparing now for the next HSK level (HSK5), meaning that I once again run into the trouble of how to study in batches. Since I haven't used Pleco's flashcard system for a while, I figured I'd use this opportunity to give this some more thought, as I want to streamline and simplify my workflow.
  • Although I do still believe that it makes sense, in principle, to distinguish between the various learning directions (listening, speaking, reading, writing), and that a high score in one direction does not necessarily imply a high score in another, in practice having 4 different scorefiles for these directions is time consuming, and as my vocabulary grows that becomes a more important problem. I don't want to just study in one direction however. As @FrancoisTaipei says, everybody is different but I think learning in one direction only is dangerous; for instance, I've met many people who are good at reading but struggle with dialogue because the sound-to-meaning link is not present in their head (they often need to visualize the word before they know what it means).
  • Having these different scorefiles has a technical downside too: it makes it hard to set things up so that, for example, words are included in the speaking test only once they have been successfully included in the listening test (although according to https://plecoforums.com/threads/min...ow-score-filter-to-specify-a-score-file.5261/ this limitation is apparently lifted in version 4.0, but the release date for that is still unknown). This problem could be addressed by using categories in a more detailed manner to specify which words should and should not be included (for instance, using the built-in Pleco functionality to split the HSK levels into categories containing 10 words per category, or use a special "Learning Pool" category), but doing again involves more work; doesn't fit the "streamlined" requirement :) (Incidentally, this was not addressed in my setup as I had described in previously in this thread either.)
  • I think that Pleco's "Alternating" subject selection is a reasonable compromise: each time I review vocabulary, it's essentially random which direction Pleco asks me (I interpret the "writing" direction as "speaking", as I do writing separately; see below).
  • Simply setting up SRS and adding HSK5 as a new category however is no good, as that would immediately introduce 1300 cards as "due" :eek:. This can be addressed using a "Include card if #reviewed > 0" record filter.
  • Of course, that then means that "truly new" cards will never come up; so I use a separate profile for "new" cards; this is useful anyway because for new cards I find it useful to study it in passive direction (char + audio) direction first.
  • Finally, I share @pdwalker 's concern that for "truly new" cards studying them once only is insufficient; so I use a third profile to practice these.
So, I've set up three profiles: Revise, Practice, and New, all of which use the same, single, scorefile.

Revise

This is the most important one. When I don't have much time, this is the only once I use, and is the one that is essential for SRS to work properly.
  • Set "Card selection/System": "Repetition-spaced"
  • Leave "Card categories" at "All categories" (since this is used only to revise cards that are due; can set this to specific categories if specific reason to only focus on some of the words that are due, but this would mess up your SRS for the other words :).
  • Enable "Card Filters/Record Filter", set "Include card if": "# reviewed", "Is": ">", "Than": "0".
  • Set "Test settings/Subject selection" to "Alternating", then enable "Test reading", "Test writing", and "Test listening"; since I interpret "Writing" as "Speaking" I then change "Configure Subjects/Writing/Show" to "Definition".
  • Enable "Commands/Repeat incorrect cards" (not essential)
  • Enable "Test Settings/Auto-play on reveal" (not essential but I like to have the audio repeated whenever possible to reinforce the right pronunciation).
Practice

If I have a bit more time, I use this to practice the words that I haven't yet got right at least 3 times in a row (in particular, this would be words that I saw for the first time today -- which I've only studied in a passive direction -- but also words that I got wrong in the review session).
  • Set "Card selection/System" to "Random" and disable "Card selection/Manage new cards"
  • "Max number of cards": "All" (can just stop the test when you run out of time or when you've answered too many words wrong and want to practice those first before trying others).
  • "Card categories": leave at "All categories"
  • Enable "Card Filters/Record filter", set "Include card if": "# reviewed", "Is": ">", "Than": "0" (i.e., consider the same set of cards that we consider for "Revise")
  • Enable "Card Filters/History filter", set "Include card if": "not correct in a row", "Is": "3" (but only consider words that we haven't yet answered correctly 3 times in a row).
  • Optionally, enable "Card Filters/Time filter", set "Include card if": "last reviewed", "Is": "within", "Amount": "1", "Unit": "Calendar day" (to practice only the words that you revised today; I enable or disable this filter depending on how motivated I am :).
  • Set "Test settings/Subject selection" to "Alternating", and set up in the same was as for "Revise".
  • Enable "Commands/Repeat incorrect words" (more important in this case because it allows you to focus on a smaller set of words by terminating a test session early)
  • Enable "Test settings/Auto-play on reveal" (again, not essential).
New

This profile is used when I feel ready to learn new words; I use it only if no words come up in "Revise" or "Practice".
  • Set "Card selection/System": "Fixed" (or "Random")
  • "Max number of cards: "All" (I then just stop the test when I feel I've learned enough new cards).
  • "Card categories": set as appropriate. This is the only profile where this matters.
  • Enable "Card Filters/Record filter", set "Include card if": "# reviewed", "Is": "=", "Than": "0" (i.e., "Revise" and "New" are mutually exclusive).
  • Set "Show" to "Chars + Pron + Audio" (I first want to have "passive" knowledge of the word).
  • Enable "Commands/Repeat incorrect cards" (not essential).
I also disable "Commands/Show statistics" for these profiles, but that's very minor.

Up to HSK4 I didn't practice writing very much; after all, it's perfectly possible to "write" Chinese on a computer/phone if you are able to "read" only. However, I want to improve my handwriting (partly because I am studying Chinese calligraphy also). Writing Chinese however truly is a separate skill so I've set up three analogue profiles "Writing Revise", "Writing Practice" and "Writing New", using a separate "Writing" scorefile, along the same lines above (except setting "Show" to "Pron + Defn + Audio" for all three profiles).

All in all, this still leaves me with 6 profiles, but my daily routine is streamlined and more efficient. Whether or not it will get me to HSK5, well, we'll see :)
 
Last edited:

pdwalker

状元
Edsko, have you considered the score based testing? If you are not familiar with it, go here (http://iphone.pleco.com/manual/30200/flash.html#subjectselectionmodes) and scroll down to the next section (unfortunately, no #href for that section) and look for the line starting with:
With “Score-based” subject selection
.

I've not used this (yet) in my testing process, but you might consider integrating it into your testing, especially if you wish to reduce the number of scorefiles you need.
 
I have, but (as far as I understand it), score based testing means that in the beginning you only study in one direction, then after the score increases a bit, only in another, then after the scores increases a bit more, only in yet another, eventually just asking you in one particular direction. That's not what I want, I want Pleco to keep asking me in all directions (with picking a direction randomly a reasonable -- I hope -- time/thoroughness compromise).
 

pdwalker

状元
Yes, I understand what you mean. I'm still thinking if I can incorporate it into my study. Even so, if you can write a character, certainly you can remember how to read it at that point?

Regarding the alternating tests - it's not random; from the documentation:
Enable Test reading / Test writing / Test listening with the three boxes on this screen, and configure them via Configure fields (which lets you choose the “Show” and “Prompt for” settings for each type). During a test, Pleco will rotate between the test types you’ve enabled for that test - if all three are selected, the first card is tested on the “reading” fields, the next card on “writing,” and so on.

I noticed this when I was using this to do a quick review of a short list. When enabled 3 tests, and my list was down to a multiple of 3, I was getting the same test for the same character every time, which kind of defeated the purpose of the enabling the multiple tests.

What I was hoping for was that each card would chose a test randomly, but over time, each card would choose each test about equally.
 
While you're right that the choice of direction is not random during the test, as long as the word selection for that test is randomized, it's the same end result: the first time you do the test, words A, B, C will be asked in direction reading, writing, listening (in that order); in the next run, words (say) B, C, A will be asked, in the same order of directions; hence, A was asked for reading first, and for listening next.

As regards "if you can write a character, you can certainly read it" -- yes, I agree (although reading fluency is a different matter, but that one I think is best addressed by reading books). But, sadly, the number of characters I can write is faaaaar far less than the number of characters I can read. I've not been focusing on that at all (and took HSK4 on the computer). So, in my alternating tests, I don't write (and instead practice reading, listening, speaking); writing I address separately because I'm so far behind.
 
I have, but (as far as I understand it), score based testing means that in the beginning you only study in one direction, then after the score increases a bit, only in another, then after the scores increases a bit more, only in yet another, eventually just asking you in one particular direction. That's not what I want, I want Pleco to keep asking me in all directions (with picking a direction randomly a reasonable -- I hope -- time/thoroughness compromise).
From my experience the default score-based testing settings (testing a different 'direction' every 200 score points) in combination with the default score adjustments using the 'rate recall quality from 1-6' option have nearly that behaviour: rating a card as 'remembered perfectly' pushes it to the next higher test direction category, whereas failing to know a newly tested direction pushes it down to the previous one, so that test directions actually alternate most of the time between consecutive reviews. The advantage over Pleco's built-in alternating mode is that the score-based approach remembers which two directions you get stuck in between: it might switch back and forth between you succeeding at knowing the pronunciation based on the meaning and you failing to write the character based on the meaning, but it will never waste your time going back to asking you to read the character (assuming that's your easiest test direction, it is for me).

I've written up a little beginner-friendly guide on how to configure a reasonable SRS test profile in Pleco, comments welcome: https://github.com/kevinstadler/pleco-flashcards/
 

Shun

状元
I've written up a little beginner-friendly guide on how to configure a reasonable SRS test profile in Pleco, comments welcome: https://github.com/kevinstadler/pleco-flashcards/

Congratulations, this looks very nice and clear! I think this could be a solution for many people. I am already at the level where I can write a character and know its pinyin/meaning after seeing it in Pleco once or twice, then repeating after a day, then after three days, seven days, then two weeks. Do you think such a rhythm is practicable with the current SRS, or would you recommend to use Random subject selection and a manually enforced schedule, with Card filters, for that? The tests I usually alternate between are Show Definition or Show Characters in the Self-graded test type.
 
Top