Whats your experience with sentence mining?

As some of you may know, I’m doing a lot more immersion this year. I’ve watched, read, and played more content in Japanese that I normally do over the past few months. One thing that immersionists seem to gloat about is sentence mining.

Over the years my sources for sentences have been iKnow, Japanesepod101, and now Bunpro. Doing a lot more reading now, I’ve no doubt come across a lot of unique sentences and vocab. I do save a lot of the unknown vocab, but not necessarily the whole sentence.

So I was wanting get some advise and expertise on those who have utilized this method. Is it effective? Is it a time consuming pursuit? Is is something that has become vital to your acquisition? What are your sources besides things like anime, manga and dramas? I really would like to know. After watching and reading Japanese content everyday, I most certainly feel like I’m learning new things and reenforcing what I’ve known, but I feel like I may need to take that next step.

You input and suggestions are much appreciated.

2 Likes

I’m not 100% sure if this counts as sentence mining, but I have this like huge list of anki decks from shows, games, etc. that I’ve seen/played. I don’t study them directly, instead if I’m looking for context on a specific word, phrase, grammar, etc., I’ll search through my decks and find cards that fit that i+1 thingy that they talk about. I’m one to think, personally, that pausing and looking for the golden nugget sentences while I’m watching seems kinda irritating.

It doesn’t take me but a couple minutes to find one, or even several, usually. And since it’s from something I’ve seen it helps a bunch with retention. I’ve most certainly started learning way faster with it than I did without it.

Hope that word vomit was remotely helpful, best of luck

3 Likes

Yeah I am not one to break out of the watching or reading in the moment to make the sentences cards, but I tend to use precompiled decks and either search for them or study them intensively for media that I am very interested in. They are not the best quality all of the time so I usually clean them up and merge them into my custom format.

It helps.

1 Like

For these pre-compiled decks, are they comprised of the dialogue in the things you watch or are they taken from other sources?

I have been at it for about 8 months now, and I would say its effective, however its hard to say how much because the overall language process is so damn long. I wouldn’t say it is time consuming because with anki addons and Yomichan it really makes the process faster. However yes there is work that needs to be done on each card depending on how thorough you want to be. I like to add the screenshot of the scene, the sentence audio, and extra definitions of other words in the sentence I dont know. Yes it takes time, but I really cant see me doing sentence mining without that extra stuff cuz it really makes the card come alive and keep me entertained. I use visual novels as my source content because its perfect for all the things I want, plus I have so many I still need to read. Only drawback ive been seeing is that sometimes I remember the scene and not the word im trying to memorize.

4 Likes

There is a community of people that have made hundreds of decks. Most of these are anime. I haven’t looked for manga ones as recently I have been more into watching to improve my listening skills. If anyone knows of any precompiled decks for manga/light novels that would be nice.

2 Likes

To clarify they are generated from the audio and the subtitle data. So generally they have an image from the scene the line of dialog and the Japanese and English sub data available for every line in the show. Some have some extras or not English and Japanese but they will say in the name usually.

The biggest issue is that some lines of dialog span two subtitle lines and they are broken in half and it can be a bit annoying getting those to line up, but those are fairly rare and not that bad.

As far as Japanese quality, they are taken directly from the subtitle data packaged with the original release so they are virtually errorless. The English is always subject to translational disagreements and potential mistakes, but that is life.

1 Like

Wow, that’s a lot. Thanks for the link.

1 Like

I really agree with @somekrazyfool
In my experience sentence mining can be a huge time sink if you want to make it look good for yourself in Anki…

Generally I have the sentence, sentence with furigana, unknown word 1 + Japanese description, unknown word 2 + Japanese description, audio (generally TTS), an image and kanji definitions. I try to have it all in Japanese, no English. I try to have all the information in the Anki card so that I would not get distracted by checking a dictionary or something.

At one point, I just copied sentences from media (eg Nihongokyoshi) and that was it… That can work well I guess if all the words / kanji are known well beforehand. (ie the +1 idea. ) . Its kinda bland for me and not fun soooo slowly I added descriptions / images to everything.

I have taken some sentences from games. Eg while playing on my Switch, I would transcribe a sentence to Anki that has a word that I don’t know and it seems important. These are the words that I remember most! haha

Web novels are great for this as its possible to copy + paste quickly. Plus there is manga / anime adaptation on a lot of the popular ones from where I can borrow an image.
Reading digital manga (eg on Bookwalker), transcribing the sentence and just using the “Snipping Tool” or “Snip and Scratch” to copy the panel can work too.
Edit: Viki or Netflix can be great for jdrama sentences.
Viki also has “Learner Mode” for some shows where its possible to see the English + Japanese / Korean / Chinese text sinked to the audio.

I do think its worth in the end.

1 Like

My experience is hugely positive. The biggest downside is time taken to sentence mine. This can be brought down to 0 with new technology.

I used to first get all the sentences in an anime episode into sentence cards in Anki, and then use an add-on called Morphman to keep track of my known words AND automatically reschedule my Anki cards so that the i+1 cards would automatically show up for reviews first. (If I ever ran out of i+1 sentences, I could just repeat and get more sentences.)

But now this process is even more seamless with a paid extension called Migaku (I promise I’m not sponsored, I just love this product so much).

Like Morphman, it also keeps track of your words, and you can use it to automatically stop a sentence mid-show if it is a 1T sentence AND the unknown word is within a certain frequency threshold (big improvement over Morphman, which would often lead you to study obscure words…) And it can be used to show two subtitles at once on Netflix.

Not only that, but with Migaku, you can simply export all the 1T sentences (whose words are within your frequency threshold) of the entire episode in bulk - no need to export the entire show before rescheduling. The process is way more seamless.

BUT…

To be honest, I don’t even make sentence cards on Anki anymore, because with Migaku, I can see in underlines the status of the words no matter what site I am using. There are three statuses: green (known), yellow (learning), and red (unknown).

My new system is to mark as learning (yellow) those red words that I searched up for the first time (and that I think I could remember next time), as green those yellow words that I understood at least one day later without looking it up, and back to yellow those green words that I forgot. This results in natural immersion for me, always using new sentences rather than hammering down the same sentences.

But I must say that learning words in context is a huge positive, and sentence mining is a great tool. Also, bulk exporting mined sentences and studying them leads to a more consistent study schedule than simply “keep immersing” and you can set particular goal amounts of new words more easily this way.

So the pro of my new system is that it’s more effective than my already pretty effective old sentence mining methods, but the pro of my old ways is that it led to more structured studying (e.g.: learn 20 new words in context per day.)

2 Likes

I’ve heard of Migaku extension but have never really used it. Even though I use Anki and Netflix daily.

Seems like a good investment, I guess I just never tried it since I wasn’t doing much immersion at the time. As of now my immersion and study time seem to be separate, it seems like you’ve combined the two. That may be the key.

I had been using https://animelon.com for a little while. Would you say that the Migaku extension is better?

It’s not either/or, thankfully. I think Migaku just recently extended support for Animelon. Either that or they are about to…

Migaku gives you Animelon-like features on Netflix.

On both Netflix and Animelon, as well as ANY web site, if there is text, it can parse it to determine the word statuses, among other things.

Additionally, it allows for easy automatic exporting of 1T sentences with definitions (and if on Animelon or Netflix, audio and screenshot) into Anki.

It actually does everything… I used to use a WHOLE hodgepodge of extensions and apps, and now it’s all consolidated.

Migaku is basically subs2SRS + Morphman + Subadub + Netflix Language Learning Extension + Yomichan + Kanji Eater + MIA Dictionary + MIA Editor while also having even more features than all these extensions combined… Oh and it works with languages outside of Japanese as well.

1 Like

I may just give it a shot. Doing daily immersion, but haven’t been retaining as much as I would like.

Thanks for the suggestion.

1 Like

The promise of animelon is so good, but the reality of it is so disappointing.

1 Like