Not all those who wander are lost

[sfuqua]

I have switched to all anki decks made from material prepared for native speakers, and I am very happy. I am making steady progress and it is very interesting.
Many of my cards are prepared from anime subs2srs decks, and they are pretty amazing. If you learn a card and you run into the utterance in your watching/listening, you probably will understand it. This can be a sort of shocking experience, since it beats the heck out of just picking up a word or two.
There are at least 2 approaches to this whole subs2srs card approach, which I might call deep and wide. If you go deep, you pick one show, or one movie and attempt to learn all the cards for it.
I am doing the opposite approach, I am going wide. I have an absolutely huge bunch of decks from many sources. Right now I have more than 300 000 cards in the deck. I have cards from 21 different anime and a few “live action” TV shows.
I am being greedy, but my favorite source of utterances is a surprise to me, the Netflix “reality” show, Terrace House. This show seems to actually be many, many hours of unscripted Japanese. It is pretty calm and normal, compared to what is generally called reality for these types of shows. It is interesting to compare the language used by this bunch of young people in casual conversation to the language used by anime characters. You always hear that anime Japanese is too casual, but the general level of politeness markers varies with Terrace House, getting very casual after the first meetings, but there is very formal language from time to time.
What is kind of shocking with Terrace house is the amount of slang and katakana English words in the language. Some of the seasons have some non native speakers, so this may affect things, but it seems to me that there are a ton of words that you will not find in the frequency dictionaries, which are actually quite common in young people’s conversations. I am sort of surprised how different the language is from “anime Japanese '', but I am also surprised at how fast I am picking it up. The whole show seems to be a bunch of conversations similar to what you get with an Assimil course.
Right now I have around 200 000 cards from Terrace House, and about 100 000 cards from a bunch of different “slice of life” anime. I also have 30 000 cards from a rather scandalous show called (in English)
“People who live off of other people's sex”. It seems to be a show about sex workers in Japan. It appears to be another example of unscripted Japanese. It was an easy 30 000 cards to add to the decks. So far it is a good source of sentences, but I am not really sure how outrageous the show actually is, because it is not available in the US (I wonder why Netflix doesn’t run it here..) There does appear to be some vocabulary in this show which is not available in my other language sources. I need to get a VPN, or find the show somewhere else. The majority of my cards are Terrace House.

I fully intend to throw away a bunch of these cards as I go along. I will never get to most of them. Many of these cards are too hard for me right now.

The traditional courses in Japanese that you find on anki take a list of sentences, or words or kanji, and have the student learn enough of these that they can start to understand enough to begin to understand enough Japanese to just take off learning Japanese through input.

Instead of working through somebody else’s list of kanji, words, and sentences, I am taking an absolutely huge bunch of sentences, and picking out the ones which aren’t too hard and learning those first. As I learn easy sentences, the harder sentences should become easier, and eventually everything is relatively easy. The way I am picking sentences is by using the ankimorphs addon, which ranks sentences by difficulty and then arranges the sentences in in an order where you never see more than one new word per sentence, but you always see at least one new word. So far it seems to be working as expected.

So far the results are just what one would hope. Not every card is good; this is the nature of subs2srs cards, but it finds a bunch of appropriate cards for me every day.

Nice.

I am happily plugging along.

[DaveAgain]

“People who live off of other people's sex”. It seems to be a show about sex workers in Japan. It appears to be another example of unscripted Japanese. It was an easy 30 000 cards to add to the decks. So far it is a good source of sentences, but I am not really sure how outrageous the show actually is, because it is not available in the US (I wonder why Netflix doesn’t run it here..)

I've got a vague memory of Japese culture having different attitutudes to sex in general.

[MorkTheFiddle]

I need to get a VPN, or find the show somewhere else. The majority of my cards are Terrace House.

I get good service from the free version of Proton VPN. It works well even though its hubs are limited to the USA, Japan, and the Netherlands. But you can connect to any url in any country through one of those hubs. When I connect overseas, it is usually to Paris, never an issue.
There is also a paid version of Proton VPN, but I don't know the benefits of using it versus using the free version.
Note that I have no affiliation with the company whatsoever.

[sfuqua]

It is about time to get a vpn going and explore. I don't even know what I am missing.
There is a lot more in subs2srs available for anime than for anything else. I have made my own subs2srs decks for some things, and it works....sometimes.

About sexual attitudes in Japan, from my limited experience, it seems to me that most Japanese people have a lot less hypocrisy and prudishness about sex that is common in America.. I find this refreshing.

Remember, there are millions of people who know more about Japan than I do. You should probably pay attention to their opinions rather than mine.

[sfuqua]

I remain ridiculously infatuated with Japan and the Japanese language.
My progress remains slow, but I continue to make slow progress.
I am very happy with my current system.
I am doing 20-40 anki cards a day. One half are from subs2srs based courses, https://ankiweb.net/shared/decks?search=Jo-Mako and https://ankiweb.net/shared/by-author/911122782 from anki and one half are from mined sentences (these days all from anime). I have the full subs2srs cards from 45 different anime shows, some with multiple seasons in my "mining" deck. This means that I have 203000 cards in total.
The object of this is not to learn all of these cards, but to have them as a resource for the ankimorphs addon https://ankiweb.net/shared/info/472573498 to search for good cards. This little program takes all of the sentence cards you have and parses them into morphemes and then calculates their frequency (or you can use a frequency file to decide this). By adding up the frequencies of all of the words in each sentence, the program calculates a difficulty number for each of the sentences. Then the program searches for sentences that have only one unknown word in them. It arranges these 1T (one word unknown) sentences in order of difficulty and moves sentences with either many unknown words or no unknown words to the end of the queue.

I love this. I am hitting a bunch of cards that have only one new word, and the sentences are slowly getting longer. Using anime sentences gives me sentences with a picture, a voice actor native speaker voice actor's voice. It sticks in the brain. Listening comprehension is getting better. At some point here after 2000 or 6000 or 10000 cards, I'm going to switch and work on just one show at a time. Or maybe just stop doing new cards on anki.

My language partner/best friend continues to be a joy in my life. She is a Chinese woman who is married and has lived in Japan for decades. We talk several times a week and we text daily. She is getting a lot more English practice than I am Japanese practice, because my Japanese production is nonexistent, beyond courtesy phrases. We never seem to run out of things to talk about. Yesterday, we were exploring our neighborhoods and sending each other pictures of wildflowers that neither or us knew the names of. Pretty random stuff. I think both of us enjoy knowing that there is a live person in another part of the world that we haven't been to.

Sometimes I think that I should have spent at least a few years in Japan. I was an ESL teacher in Asia, so it is kind of strange that I never lived and worked there. I wasted a few years in different places where I wasn't really having new adventures, and I could have spent, by my calculations, 3-4 years in Japan and still been to the Philippines in time to marry my wife and have my current family. I talked about this with my friend and she laughed at me. She said, "You would have never lasted in Japan, some Japanese girl would have snatched you up, and you would be gone..." I guess Japanese girls are dangerous. She is married to a Japanese guy, so I guess she knows how it goes. Ah, the road not taken...


I mentioned my daughter in an earlier post, and she has made some decisions. She is going to continue college and adjust her major as she goes along. But the big news is that she has a professional (low paid, gig work) job with one of the local theaters that does plays and musicals. She will be doing a bunch of different offstage things with the company, lighting, stage managing, sound, etc. it isn't much, but it is a foot in the door. She says that realistically she probably has more talent in design than she has in singing, and she loves musical theater and just wants to be a part. She assesses, realistically I think, that she has a great high school voice, but not a one-in-a-million voice. Anyway, she can do this backstage job while she continues as a college student for the next few years.

It sounds like the kind of vague plan that is appropriate for an 18 year old.

Please forgive my awful typos. I hate to admit it, but my vision is affecting my ability to notice misspelled words. I've always been sloppy, but this is getting ridiculous these days. I'll try to do better

I am happy and I continue to plug along..

Just fixed some typos and added a sentence that I forgot to put in.

[sfuqua]

In case anybody is thinking about using ankimorphs to filter out good sentences to review, there are some problems:

1. When you first start working with ankimorphs you will get a bunch of sentences like, "Wha-", "Huh?" "yes", "Yeah", "OK", and the like. There are ways to avoid this, but the easiest thing to do is just to power through, After a 100 cards or so, you start to see better sentences and vocab. These "bad" sentences are the result of algorithm. The words are common, and the sentences are short, so ankimorphs picks them first. Each day, after your reviews, you run ankimorphs and recaculate everything. Ankimorphs moves other sentences with these words to the end of the queue because you already know these words.
2. For Japanese, the mecab parser that is part of ankimorphs makes mistakes, particularly with words that are in kana. It may just give up and declare that a sentence with several words is just one really, really long word. My mecab/ankimorphs setup thinks that やっぱりかずさってなんでもできちゃうんだね is Two words. There are several words in there, or course, and some of them I don't know. This is about a 5T sentence, not a 1T sentence.
3. In a related problem, the subtitles to anime are not always the best. If you stick to quality subs2srs decks, this may help (decks by Jo Mako are well made in general), but the main problem is with the original subtitles. There is an absolute ton of kana, and it might be easier if there were kanji, and the structure of the sentence was clearer. I hope that there will be more kanji as I go along. When I compare the language in anime to the language in newspapers, well it is quite different.
4. Some of the sound is messed up on some cards. It is sad to have an interesting sentence with an interesting picture and have the sound cut out halfway through the sentence. While the usual solution should be to just delete the card, you can always replace the sound with a tts file. Anki and ankidroid now have the same setup for tts. Just putting {{tts ja_JP:VARIABLE}} will cause anki to pronounce VARIABLE in Japanese for both ankidroid and Linux. I think it will work on other operating systems too. It is very convenient if you want to replace faulty sound on a given card. You can just deleting the link to the faulty audio file on the anki card, and make sure that there is code on the card to use tts to say the variable if there is no sound file.

Of course tts voices are different than native speakers talking, To my ear, tts voices are generally easier to understand.

The longer I work with ankimorphs, the better it seems to work.

[emk]

I am following your Subs2SRS experiences with great interest! You're the other person on the forum who uses it very heavily, and I'm very interested in how the techniques work for other people. If you aren't following my Spanish log, you might find an interesting contrast or two there.

3. In a related problem, the subtitles to anime are not always the best. If you stick to quality subs2srs decks, this may help (decks by Jo Mako are well made in general), but the main problem is with the original subtitles. There is an absolute ton of kana, and it might be easier if there were kanji, and the structure of the sentence was clearer. I hope that there will be more kanji as I go along. When I compare the language in anime to the language in newspapers, well it is quite different.

I think I've mostly solved this for Spanish (and probably French), at least for easier "intermediate" TV shows. The various Whisper-based transcription services produce largely accurate transcriptions and OK-ish timing data for sufficiently popular languages. Whisper also has a feature where you can give it a "prompt", which is really just a sample of what you want the output to look like. This might be useful for something like controlling the mix of kana and katana, but I haven't tried it.

MacWhisper is supposed to be quite good. for people with access to a Mac. So if you wanted to tackle a particular show at some point, this might be worth a look.

Cost for using the OpenAI API is about US$0.01 per minute of video. Unfortunately, none of the defaults in my "substudy" tool are really tuned for producing Japanese transcriptions. It deals much better with languages that use spaces. There is a "raw" mode than returns Whisper SRTs with minimal post-processing that might work for Linux users, but I wouldn't recommend trying it right now.

4. Some of the sound is messed up on some cards. It is sad to have an interesting sentence with an interesting picture and have the sound cut out halfway through the sentence. While the usual solution should be to just delete the card, you can always replace the sound with a tts file. Anki and ankidroid now have the same setup for tts. Just putting {{tts ja_JP:VARIABLE}} will cause anki to pronounce VARIABLE in Japanese for both ankidroid and Linux.

Oh, that's terrific! Anki templates also support a "conditional" feature. So you could add a blank "UseTTS" field to your note, and do something like:

Code: Select all

{{^UseTTS}}{{Sound}}{{/UseTTS}}
{{#UseTTS}}{{tts ja_JP:JapaneseText}}{{/UseTTS}}


Then when you hit a card with broken audio, you could just use AnkiDroid to set "UseTTS" to "y" and continue merrily on with your reviews. (Actually, this seems like something I should somehow support with substudy. But I also have access to the OpenAI TTS, which is really good. Hmm.)

The longer I work with ankimorphs, the better it seems to work.

I am absolutely fascinated by the AnkiMorphs. I never really needed anything like this, because the jump from French+English to Spanish was much smaller, and I could basically brute-force it with the first 100 cards and a grammar quick-reference. Also, I'm way more willing than many people to tolerate what Ankimorphs calls "multi-target" sentences. They make my life harder, but not impossibly so.

Really, Japanese presents a whole host of challenges I don't really understand. Middle Egyptian gives me some hints, but there, I have Assimil to introduce stuff in a reasonable order. And Egyptian's writing system is much simpler than kanji—you realistically only need to learn about 60 phonetic signs, and another 50 or 75 determinatives. But determinatives are easy; they're used to disambiguate the phonetic signs. The determinative for "sandal" is literally a picture of a sandal, for example.

So I'm very interested in seeing all the tools and techniques you're using to study Japanese with audio cards. Thank you for describing them!

[sfuqua]

Things are still going well, but I have started to notice some resistance to study lately, which makes me think that I may need to reduce my new cards soon.

But I'm still keeping up my 30 new cards a day pace.

I talked a little bit last time about how to make half broken subs2srs cards work. Often the English subtitles for a given card won't match the Japanese very well, or at all. Sometimes it is bad translation, sometimes it is a difference in culture, and sometimes it is just bad subtitles. Some of them can seem like something out of "What's up Tiger Lily? https://www.youtube.com/watch?v=_f20pv9QBc0" Anyway I always put a google translation of the Japanese on the back of the card, as a hint in case the English doesn't make sense.
Rather than doing a big download and upload of 200 000 cards, I just use an addon to put translations on the next 100 cards I will reach in the deck. Since I'm only doing 30 new cards a day, this means that I always have that translation there. The addon is at https://ankiweb.net/shared/info/1536291224 .

I continue to plug along, although the last few days it has been getting harder.

I assume anybody who wants to see the Shogun has already seen it, but if you haven't, the following contains spoilers.
One of the fun things I have been doing lately is watching the 10 episode show Shogun, which isn't teaching me much Japanese, even though it is 90% in Japanese. It is too fast, there is too much unknown vocabulary, and the language used is archaic. I am a fan of the Shogun; I first read the book 40 years ago, and I reread it recently when I heard that there was a new miniseries going. Reviews have been good and most of the criticism seems to be based on comparisons of the current miniseries to the actual history, the original 1970s book, or the 1980s miniseries.
Some people just don't like films with cross cultural romances or a show with swords.

It is fiction, so it differs from the history quite a bit in places; I like most of the changes from the book, and I am married into a cross-cultural romance, so I don't mind those.

There is a lot of good acting in the show, but I think that Anna Sawaii is perhaps the most amazing. Anna is a lovely young woman, but it is her acting chops that leave the greatest impression.

Here is Mariko (Anna's character) translating for a man she is very attracted to when she is ordered to translate for him during a visit to a courtesan. The courtesan senses exactly what Mariko and Anjin (the man) are feeling, and creates some opportunities for, uh, interesting translations. https://www.youtube.com/watch?v=36TCRoNJrOg . Anna has many other great scenes in other parts of the show

Anna Sawaii is a treasure and I can't wait to see what she does next.

My real celebrity crush from the show is another amazing young actress, Moeka Hoshi . She has a smaller role in the show, but is a constant scene stealer. Somebody made a compilation of some of her scenes from the show, Look at her eyes https://www.youtube.com/watch?v=JJ4wAmCr1wo .

One of the things that I noticed about the show is the importance of nonverbal communication, especially with eyes. It reminds me of Samoa in many ways. When things can't be said out loud, they get said with eyes.

I thought that my language buddy in Japan had the best comment on the show that I have seen. She said that Japanese people wonder why Americans like it so much; it is a very Japanese story and production.
I asked her, "What do you think of it?"
She said, "It isn't as funny as 'The Big Bang Theory'."
That certainly is true.