Sentence Mining Japanese: A 5-Step Workflow to Read and Remember (2026)
TL;DR: Sentence mining is the practice of saving Japanese words you meet while reading or watching, in the entire sentence they appeared in, and reviewing them on a spaced schedule until they stick. The 5-step workflow is: choose a source you actually want to read, hover any unknown word, save it in one click, review on a spaced schedule, and track what's sticking. You can run the full loop with immit (free, no account, hover plus a built-in spaced repetition system) or with Yomitan plus Anki if you prefer to assemble it yourself.
Key facts:
Sentence mining works best for early-intermediate Japanese learners and above.
A complete sentence mining workflow has 5 steps: choose source, hover lookup, one-click save, spaced review, track progress.
Daily time commitment is 10 to 20 minutes.
Recommended new cards per day is 5 to 10 for beginners, 10 to 20 long-term.
Tools to run the loop: immit (no setup, free, no account) or Yomitan plus Anki (more setup, free).
Expected outcome after 6 months of consistent practice is reading short articles without hovering on most words.
I was reading NHK News Easy on a slow Sunday morning when I noticed something a little embarrassing. I'd looked up the same word three days in a row. Not a rare word. Just a word my brain refused to keep. I'd hover, read the definition, nod, and move on. Three days later, same word, same hover, same nod.
That loop is what mining sentences fixes. You stop looking up the same word forever and start saving it the first time, reviewing it a few times across the next few weeks, and meeting it again in real reading. After a few cycles, the word is just yours.
I'm Ben, an American intermediate-advanced learner of Japanese, and I co-build immit with my Japanese-native wife Serena. This post is the version of the explanation I wish I'd had the first time I heard "sentence mining" in an immersion-learning thread. If you've never mined before, what does the loop actually look like, and what's the smallest tool stack that runs it?
Sentence mining is the practice of extracting sentences from immersion content (manga, news, video subtitles, social media) and adding them to a spaced repetition system, so you study vocabulary in context rather than in isolation.
Sentence mining works best when you focus on sentences where you understand all but one element, whether that one unknown is a word or a grammar point. This i+1 principle prevents overwhelming the memory with too many unknowns at once, and it's why mining outperforms cramming bare word lists for long-term retention.
Two things make sentence mining different from regular flashcards. First, the target word comes from content you chose, not from someone else's word list. That means every word you study is a word you've already needed once and will probably need again. Second, the entire sentence comes with the word. When you review the card, you don't see "新聞 = newspaper" floating in a vacuum, you see the actual sentence you were reading when you met it.
Learning vocabulary in context enhances retention because each word arrives with multiple associations and hooks for memory, making the word easier to recall later. A single word can carry multiple meanings depending on the context, and the saved sentence captures the specific meaning you encountered. Japanese sentence structure (particles, word order, the verb at the end) becomes familiar to you sentence by sentence rather than as a separate grammar exercise. Recall gets faster because the brain isn't reaching for an isolated dictionary entry, it's reaching for a memory.
The result for early-intermediate learners is that vocabulary growth tracks reading instead of running parallel to it. The hours you spend on language learning become the hours that build your word base.
The first step in sentence mining is picking a source that is comprehensible enough that you can follow most of it, and interesting enough that you will come back to it tomorrow. Mining lives or dies on whether you come back tomorrow.
How to find sentences worth mining
Two qualities matter. First, the content should be roughly comprehensible. Apply the i+1 principle: pick sentences where you understand all but one element, whether that one unknown is a word or a grammar point. If long sentences come at you with five unknown words each, the brain gives up and you remember none of them. Second, the content should be something you'd read anyway, even if Japanese weren't involved. A manga you've already read in English, a Twitter account in your hobby, YouTube videos from a creator with Japanese subtitles, a recipe site, a fan wiki for a game you play.
Good first-month sources for early-intermediate learners include NHK News Easy (short, with furigana), Satori Reader (graded, with audio), simple slice-of-life manga like Yotsuba&!, and Twitter accounts in topics you already follow in English. The "comprehensible enough plus interesting enough" combination is what sustains the habit. Skip sources that are too hard for now; they'll still be there in six months.
Picking new words to focus on
You don't choose words ahead of time. You let the source choose them for you. Whatever native speakers happen to use in the content you picked is the list. That's the whole point of finding new sentences in real material instead of memorizing pre-made decks: every word that goes into your queue is a word you've already needed once, in a real context, with a sentence behind it. The "list" is just the union of every unknown word you've met across everything you've personally chosen to read.
The second step is to hover any word you don't recognize and read the meaning instantly, without leaving the page. A popup dictionary is the tool that makes this possible.
The lookup step is where most beginners give up on mining, because copy-pasting a word into Jisho or Google Translate every few seconds breaks reading flow completely. Browser extensions streamline the process of looking words up and creating flashcards from digital content, letting you mine sentences without interrupting your immersion. The fix for the lookup step specifically is a popup dictionary that surfaces the reading and the word's meaning the moment you hover, without leaving the page.
Two main popup-dictionary browser extensions run this step today. Yomitan is the free, open-source choice that the immersion community has used for years. It's powerful and customizable, but it's designed for people who enjoy assembling their tools, so the install includes downloading dictionary files, configuring keybindings, and tuning popup behavior before the first lookup. Plan an hour the first day if you go this route.
immit is the no-setup browser extension alternative. Install the extension, hover, and you have a reading, definition, pronunciation, and example sentences. The dictionary is bundled, works offline, and covers JP-EN, EN-JP, EN-EN and JP-JP directions. When you report a mistake from the popup, a person reviews and corrects the entry, so the definitions improve over time.
What hover lookup gives you that copy-pasting doesn't is speed and context. Reading flow stays intact. The word you just hovered is still in the sentence in front of you when the popup closes, which is what makes the next step (saving) feel weightless instead of heavy.
Try immit free. Hover any Japanese on any webpage. No account.
The third step is to save the word as a flashcard with one click, capturing the entire sentence as context. A good sentence card always carries the original sentence with it.
What goes on a word card
A good word card captures the target word and the surrounding dictionary information automatically: the reading, the word's meaning, the part of speech, the pronunciation, and pre-curated example sentences for the word. Together these turn an unknown word into a card you can actually review later. Some miners also paste the original sentence they were reading onto the card by hand, because the situation you encountered the word in becomes a useful memory hook later. The card structure in any tool is flexible; the rule is that the more weightless saving feels, the more often you save.
In immit, save is one click on the popup. The target word, the reading, the word's meaning, the part of speech, the pronunciation, and example sentences from the dictionary go onto a flashcard automatically, together with the sentence you were reading and the page it came from. immit creates cards for you in the background; you stay in the reading flow. A Yomitan plus Anki setup with AnkiConnect can do the same once you map Yomitan's {sentence} field to your note type; it takes longer to assemble (downloading dictionary files, configuring keybindings, tuning popup behavior, installing AnkiConnect, designing a card template, mapping Yomitan's field markers) and is the option for learners who enjoy that configuration work as part of the hobby.
Whichever tool you create cards with, the rule is the same: save the first time, not the third time. Trust that the review system will handle the rest.
The fourth step is to review your saved cards on a spaced repetition schedule. A spaced repetition system is a learning technique that schedules each card to reappear just before you would otherwise forget it.
Spaced repetition systems use algorithms to determine the optimal time intervals for reviewing flashcards based on how well you know each card. The algorithm adjusts review frequency to your individual performance, which prevents overwhelming review backlogs and lets you study vocabulary in context without the daily pile exploding. Get a single word right, and the next interval grows (next time you'll see it in three days, then a week, then three weeks). Get it wrong, and the interval resets so you see it again soon. Over a few cycles, the words you review move into long-term memory and the daily review pile stays manageable.
Why sentence mining works
Sentence mining works because the spaced repetition system is doing the memory work in the background while you do the reading work in the foreground. You're not memorizing word lists in a vacuum; you're recalling words that came from your own mined sentences, with the context attached. Recall on a sentence card is faster than recall on a bare-word card because the sentence carries the meaning along with it, and the multiple memory hooks (the situation, the surrounding words, the source content) all help bring the target word back.
In practice, plan on 10 to 20 minutes a day. Anki cards from yesterday's reading enter the queue automatically. Old cards come back when their interval is up. A typical session is 80% words you remember (those leave the daily queue and come back later) and 20% words you almost remember (those come back tomorrow). Modern schedulers like FSRS, which most current spaced repetition system tools use, are calibrated to keep the load near that ratio without you tuning anything.
immit's daily review is built in, runs offline, and uses FSRS out of the box. Anki is a popular free, customizable tool for language learners that uses a solid SRS algorithm; if you already use Anki, that loop also works fine. The choice between Anki and immit comes down to whether you want to assemble the setup yourself or skip it. Both use FSRS.
The fifth step is to reread old material once a week and notice the words you no longer need to hover. That observation is the feedback loop that confirms the workflow is working.
The final step is the one almost everyone skips, and it's the one that tells you the workflow is working.
Once a week, go back to a source you read the previous week and reread a few paragraphs. Read the same sentences a week later and notice the words you don't have to hover anymore. Those are the words mining moved from "vaguely familiar" into "just part of the language." Seeing it happen is what keeps the habit going on weeks when motivation is low, because you have direct evidence that the time is compounding and the process gets faster as you read more sentences.
Tracking can be as light as you want. immit shows how many words you have saved and how far along each card is, so you can watch the curve grow without manual logging. A small notebook works too. Even just paying attention while you reread is enough.
The compound is the point. If you save 10 words a day for six months, you've added roughly 1,800 retained words you actually saw in real content (assuming typical SRS retention around 85-90%), attached to the mined sentences that put them there. That's enough to drop your hover rate sharply on familiar topics — slice-of-life manga, simple blog posts in your hobby, easy news rewrites like NHK News Web Easy.
Start sentence mining from early-intermediate up. If you can read hiragana, katakana, and a few hundred kanji, recognize a few hundred common words, and understand the rough shape of basic grammar (one grammar point at a time, particles and verb conjugation broadly familiar), you're ready as a language learner. Start with short immersion content (a paragraph at a time is fine).
If you're earlier than that (still learning the kana, no grammar foundation yet), mining will frustrate you because almost every word in every sentence will be unknown. A foundations course or a structured beginner app for two or three months first gets you to the level where mining starts paying back.
After about six months of consistent daily mining (10 to 20 minutes a day, 10 to 20 cards), most language learners report reading short articles without hovering on most words. That's the milestone the workflow is built around.
What's the difference between sentence mining and just using flashcards?
Sentence mining differs from regular flashcards in two ways: the source content is your own choice, not someone else's pre-made deck, and the original sentence is saved on the card alongside the word. Learning vocabulary in context enhances retention because each card arrives with multiple memory hooks (the source, the surrounding words, the situation), which makes the word easier to recall later than an isolated dictionary entry would be.
Do I need Anki for sentence mining?
Anki is one popular tool for the review step, but it is not required. Anki is free, customizable, and uses a solid SRS algorithm, so a Yomitan plus Anki setup is a complete mining stack if you want to assemble it yourself. immit runs the full loop (hover, save, daily SRS review) with no account and no setup, and is enough on its own. Pick the tool whose setup cost you're willing to pay; the workflow is the same either way.
How many cards should I mine per day?
Start with 5 to 10 new cards a day for the first two weeks. That cap is enough to build the habit without growing the daily review pile faster than your retention can keep up. Once a daily session feels routine, 10 to 20 new cards is the long-term sustainable range for most early-intermediate learners.
Can I sentence mine from Netflix and YouTube?
Mining from video subtitles is a common variation of the workflow with the same five steps. The only difference is that the source in step 1 is a subtitle line instead of a webpage paragraph. We'll cover the video-specific setup in a separate post on mining from Netflix.
Read something you like, hover what you don't know, save it in one click, review for 15 minutes a day, watch your reading get easier. That's sentence mining. Six months in, your new vocabulary will be measured in the thousands, every word attached to a sentence you actually read.
If you want the loop running today with no setup, install immit free (Chrome Web Store, no account). For readers who want a deeper dive into the Anki side specifically, our companion post on building an Anki mining deck covers the card-template and AnkiConnect details.
