Postby raikiro » Wed Jun 30, 2021 5:24 pm
The end of a month and of a quarter year, let's see what happened.
Starting off with Japanese:
In my last post I wrote about my issues with anki. Changing the intervals did not help too much, because I have a rather high fail rate, leading to too many lapses. I have therefore decided to change the lapse intervals. Ankis default options has it at 0%. I had already increased it to 50%, because it does not make sense that you would always need to start from 0 with cards you are already familiar with just for failing them once. I saw a post recently about how you should set your lapse interval to 100%, so that the intervals can never get shorter - after all, if you've already managed to remember a card after x amount of days, it shows that you brain should remember them sufficiently well. After that, I set my lapse interval to 90%. Meaning, if I fail a card repeatedly (on the same day) the interval gets shorter, but just failing it once does not set me back too much. The result: after about 5 days of ankiing with the new lapse interval, my cards/day have decreased from about 230 on average to about 150.
The anki simulator addon (which calculates based on your settings and your retention and lapse percentages) shows the reviews decreasing further in the next couple of weeks. If I run the addon with the old lapse interval of 50%, it predicts the number of reviews to stay the same for the foreseeable future. So right now this seems to be my best bet to finally get those numbers down.
Is this the solution to all of my anki problems? Probably not. It is probably not optimal, because the frequent lapses show that it is not working the way it should. But considering that I've been doing anki for months now, and no real improvements have been made, anki is not the solution. The solution is more exposure and reading. Which is what I wanted to do anyways, but I was lacking the time. Spending ~50 minutes a day on anki left me not wanting to "waste" any more time with Japanese, while my main focus should be on Spanish.
The solution would probably be just to get rid of anki. And rely on that I will get exposed to all the characters I need if I just read enough. But since I like reading in multiple languages, I am afraid I would forget large amounts of kanji when not using Japanese for a while, only to have to re-learn them every time.
So this is my attempt so fix it. Reduce my reviews further, and instead use the gained time to read more. Already the time that I spent on my flashcards daily has gone down quite significantly, and I have started reading a bit every day.
Another change I made recently is that after lapsing I wanted to get the word correct twice. I thought that if I saw it more frequently after failing I would remember it better. But after getting it correct the first time, the chance to fail it the second time that day is only 3%. This means two things - firstly, seeing it a second time did not, in fact, increase my overall retention and and decrease my reviews and secondly, considering that I almost always get it correct the second time, if it does not help me remember it the next day, I'm wasting my time. So I have removed the second lapse step again.
After having read the Harry Potter series, I started looking for other interesting fantasy novels, preferably ones that I am already familiar with. Unfortunately, whilst the big series such as Twilight (my go-to besides Harry Potter) or the Hunger Game series have been translated into Japanese, they only seem to be available in print. And importing it would be very expensive, plus I prefer having a digital version of the text anyways to read on my kindle or with the aid of other reading tools. And even with books that have digital versions, they are tough enough to find online. You could try to use amazon.jp with a fake address - but even for that you need a credit card, which I don't have (they are not common in Germany). I have found honto.jp, a page that explicitly allows foreign customers to order books and I might try it soon. However, from the series that I am interested in that I already know, I have not found to many results there.
So I decided to forget fiction for a moment. Many people say that fantasy novels are much more difficult than non-fiction texts anyways. Last week I bought 日本人とは何か。(上巻)神話の世界から近代まで、その行動原理を探る, which was available through the German amazon. I've read through the prologue and so far I don't feel like it is easier than novels. Maybe it gets better.
No statistics to report, as I did not listen nor finish a book.
Onto Spanish:
This month was definitely more listening than reading, which is very uncommon for me. I only finished reading one book, and two audiobooks. I started reading more, but like I stated previously, I will only count the words I've read when I'm done with the book.
I started working through the Spanish without toil I mentioned last month. The beginning is very easy, and I've worked my way up to lesson 78 (of 111) in the last weeks. It is not particularly challenging yet, but the active wave is taking quite some time, so I will probably be slowing it down a notch.
My approach to the passive wave: 1. listen without text 2. listen while reading 3. read out loud, check unclear words 4. shadow x3 5. read out loud again. The most difficult is the initial listening without text. I often find myself struggling to understand, but when I look at the text, it makes perfect sense. Considering that I am more of a reader than a listener, this is unsurprising. And then I sometimes struggle with the shadowing. Because I have no speaking practice I'm struggling to produce the sounds quickly enough, which is why I go through it several times in the lesson, and I plan on going through it again. I have used the programm audacity to remove the gaps of silence in between and created a file with all lessons after another, so I can listen to the entire course to review.
For the active wave: 1. read the Spanish text aloud once 2. cover the Spanish and, using the translation, try to write the Spanish text in my notebook 3. correct the text I came up with using a different color pen 4. take sentences that I struggled with or made weird mistakes in and add them to anki.
So far, this has been going great. I decided to make it a bit easier for me by reading though the Spanish text once first. Currently I am planning on perhaps doing another active wave later, then without first looking at the Spanish, and see if it helped. If could then use my current notebook to compare the mistakes and see my improvement (hence the different colors).
Statistics:
hours listened: 34,5
words read: 32.537
Books I finished in June:
Harry Potter y el prisionero de Azkaban
Harry Potter y el cáliz de fuego
Cautivas del berserker book 6
(21/52)
Currently working on:
Sinsajo (11 chapters in)
El beso carmesi (11 chapters in as well)
Moving onto reading tools.
I have now had some more time to use readlang, and I checked out LanguageCrush (formerly OPLingo) as well.
My criticism of readlang I already posted last month, so I will mostly write about LanguageCrush now. Since I have a lot of thoughts, I will use bullet points to try to keep it short.
- No option to import books directly -> need to convert into text document using calibre (I don't mind though)
- Passage length: with readlang, you have the entire book in one. The first time I imported something into LanguageCrush, it split the book into parts, or "passages" of 2000 words. Very annoing. I now imported some more: one book in Spanish, Russian, and Japanese. The Spanish and Russian ones were split into parts of about 7-7.500 words, which is fine with me. Having a book be split into a handful of parts is not a problem. Having it split into 150+ parts, like in LingQ, is an annoyance. The Japanese book that I imported, however, was split into only 4 parts, with about 25.000 words each - so they are very long. Perhaps this has something to do with Japanese being a beta language, or with the characters. I don't know, it just seems weird that there is such a huge difference.
- When importing a passage, the box for "share with community" is automatically checked; I always worry that I will forget unchecking it and getting into trouble for sharing books
- Hovering over a word shows the translation pretty much immediately, which is very convenient for quickly following an audiobook (make sure to have "display passages on one page" unchecked though, otherwise it is very slow)
- But: there seems to be an issue with RAM usage. Every once in a while the page would get very slow and start lagging. I noticed that at one point that once browser tab was using more than 2 gigabytes of RAM. Probably a bug? Closing and re-opening fixes this, but it happens every once in a while.
- It has auto tts, which I like
- You can split and link words, which is perfect for Japanese - LingQ would always assume, sometimes wrongly, which characters made up one word. Here I can choose for myself.
- You can not change the font size! LanguageCrush uses my monitor space much better than LingQ with its big borders, so the amount of words shown is better. However, the font size is much bigger than I would need. Zooming out in the browser makes everything smaller, but I would like an option to show more text on one page instead.
- There does not seem to be an option to add several passages to a course at once, by marking them. I imported one book without telling the tool to add everything to one course, and then had to manually edit every single part to add it afterwards.
- There is no clear indicator, how many words are in the passage. It only shows the amount of unique words, which is obviously lower. I'd like to tell with a quick glance how long the passage actually is.
- There is no switching between languages. You mark the language the passage is in, but there is no drop-down to choose which language you want to see like you have with LingQ or Readlang. Everything is mixed together. Therefore I am not sure if the "known words" are counted separately for each language as well, or if there is overlap. I would hope not.
- It color-codes unknown words and words that you are learning. There is a button at the end of a passage to mark all unknown words as known, which I will be using a lot, especially in the beginning, as the tool first needs to find out what I know. Considering I have quite long lessons (especially in Japanese) I would like an option to mark all unknown words until a certain point as known. By using a bookmark or something maybe?
- The colors are quite intense. Probably the biggest plus of Readlang is its minimalistic and clean look. In LanguageCrush unknown words are red and words you are learning are an intense dark green, which is a bit too aggressive for my preference. The light yellow LingQ uses is more pleasant to the eye.
This last point I'm not even going to try to put in a bullet point. The free version has a limitation on word and phrases lookups. This only goes for words you click: just hovering over them will always show the translation. Now, naturally, before I pay for something I like trying it out, and I have encountered tons of issues over these limitations.
If you have reached your 15 word lookup per day, you can not:
- Split/link words: in Japanese, this is horrible; if a word is split incorrectly, you can not correct it
- Mark words to be able to copy them for checking an external dictionary
- Set bookmarks (why not? I don't understand..). It opens the word lookup on the side, and then tells you that you have reached your daily limit
- Everytime the popup about how many lookups you have left comes up, it deselects words that you previously selected - even if phrase lookups are not yet exceeded
- Save/change translations: extremely bad. You can see a translation that is completely wrong, of a wrong reading of characters, and you can not correct it. You can save the word as "learning", but with the wrong translation
The issues relating to the word lookup limits can be removed by purchasing the premium plan. And at 7$ a month, this is cheaper than LingQ. I can definitely live with this. Some of the restrictions however, such as the bookmarks or being unable to correct translations, just seem bad. Saying that you can use the tool completely for free just isn't true. (Unless you only read one page a day anyways.)
Don't get me wrong, I am more than willing to pay for a tool that works. But some of the other issues and bugs are very annoying. I understand that this is a rather new project, and that the developers are working on it actively. So there is a good chance some of these will be fixed in the future, and more customization options (such as font size and color intensity) will be added.
So how do I move on?
I am pretty sure that I am going to cancel my LingQ subscription. The community courses are still available to free accounts, and the whole "losing your saved LingQs when you cancel" really tries to force you into staying. I would say that I got my moneys' worth out of it, but it might be time to move on. You can export all your LingQs and presumably import them again later on, which might be a workaround. However, it just spits out an excel document of everything - I've tested for Russian and got over 17 thousand LingQs, not sorted by known or learning. So I'm not sure how useful this would be.
Readlang has proven to be a very simple, clean looking, and effective tool. I would like more statistics such as known words though, and the fact that the word popups do not have Furigana makes it pretty much useless for quickly readingJapanese. You always have to wait for the dictionary to pop up. And since it is pretty much a dead project, no more work is to be expected. Having the entire book in one document is great - but without stuff like statistics, it is just a tool were I can click words to get their meaning - and for that, why not just use my kindle?
LanguageCrush seems to be the way to go for me here. Being able to customize words in Japanese is huge compared to LingQ, and the longer passage length is great. I just hope it stays the way it is right now, and does not decrease again. Plus, there is hope that the people continue working on it. I also still have to try out the android app.
The LingQ challenges were one of the reasons why I decided to break the year into quarters with a designated main language. I have not maintained my LingQ statistics, instead keeping track of everything in a separate spreadsheet. Therefore, this is not really important anymore.
I still liked the idea of having one focus for a set amount of time though, as it prevents me from wandering around between my languages too much.
How do I continue? I still have at least one quarter I wanted to spend on Russian this year. But I still have two Spanish book I started that I want to finish (and since one of those is from a series, I might want to finish the entire series). Plus, I should absolutely stick with Assimil for now. I don't want to drop it now that I'm starting to get somewhere with it. So, no Russian just now.
Current idea: finish Assimil Spanish, and continue focusing on Spanish reading in general. Once I finished Assimil Spanish, it might be time to get back to the two Assimil Russian that I have lying around. In the 1971 version I got to lesson 47 (of 100), and in the 1991 version I got to lesson 34 (of 71). This means that I am pretty much halfway with both of them, just a bit away from the start of the active waves.
Getting into reading Russian is much more difficult that it was for me to get into reading Spanish, so I feel like finishing the Assimils would help with that.
So, not really a clear goal this time around. I will just see how it develops over the course of the month.
3 x