Sentence Mining

Also known as mining, i+1 sentences, 1T sentences, targeted sentence cards

Pull sentences you almost understand out of things you're already watching or reading, and turn each into one flashcard.

Sentence mining is a card-making discipline rather than a whole method. Instead of studying a pre-made list of words, you take your flashcards out of the material you personally consume: a line from an episode, a sentence from a novel, a phrase from a podcast. The governing rule is that a mined sentence should contain exactly one thing you don't know — the community calls this an i+1 or 1T ('one target') sentence. Everything else in it should already be familiar, so the sentence itself supplies the context that makes the new item memorable.

The argument for it is that words are not learned in isolation. A card that says 'kuyashii = frustrated' teaches you a gloss; a card carrying the line a character actually said, in a scene you remember, teaches you the register, the collocation and the grammar it sits in at the same time. It also guarantees relevance: everything on your cards is by construction something you met while doing something you wanted to do.

Mechanically, modern mining is fast because of tooling. A pop-up dictionary over a subtitle track, or a reader with one-click export, will build a card containing the sentence, the target word's definition, the audio clip and often a screenshot, in a couple of seconds. Card formats vary: some people put the whole sentence on the front and test comprehension, others cloze-delete the target word, and a well-known minimal-pairs variant tests only the one item with the sentence as a hint.

A realistic week is: watch an episode with subtitles on, save six to ten lines that were nearly clear, spend ten minutes tidying the cards, and review the accumulated deck daily. Practitioners generally cap new cards hard — ten or twenty a day — because the review load compounds and mining is very easy to overdo.

It is worth being clear about what sentence mining is and isn't. It is a way of feeding a spaced-repetition system with better material. It is not a source of input, and it does not replace the hours of watching and reading that generate the sentences in the first place. Mining without immersion produces a beautiful deck and very little comprehension.

How it works

Consume native or near-native material you actually enjoy — that is the ore.
Save only sentences where a single element is unknown; skip anything with two or more.
Build the card with the full sentence, the target item's meaning, and the audio if you can get it.
Keep one target per card, and delete cards you no longer care about rather than fighting them.
Cap new cards per day (ten to twenty is common) so reviews stay under half an hour.
Review daily in an SRS and re-encounter the words in the wild — the card is a reminder, not the learning.

Origin Khatzumoto and the AJATT community, late 2000s — Popularised on the All Japanese All The Time blog as the way to feed Anki, then systematised by the Japanese-learning community and built into mining tools.

Good for

Intermediate learners drowning in near-known vocabulary
Retaining words in the grammar and register they actually appear in
Anyone using subtitle tools or pop-up dictionaries, where mining is nearly free

Trade-offs

The i+1 rule is much harder to satisfy than it sounds. Below a large passive vocabulary almost every native sentence has two or three unknowns, which makes mining a poor fit for beginners — a frequency deck is faster at that stage.
It scales badly if you don't cap it. Every episode offers dozens of candidate cards, and unbounded mining produces a review queue that eventually eats the immersion time that fed it.
Card-making is a satisfying, measurable activity that feels like studying. It is easy to spend more of your week curating a deck than listening to the language.
Recognising a sentence on a card does not mean you can produce the word, and mined decks skew heavily toward passive recognition unless you deliberately add production cards.
Direct research on sentence mining as such is thin; the strong evidence underneath it is for spaced repetition and for context in vocabulary learning, not for this particular workflow.

Who advocates this

Khatzumoto Creator

Popularised mining sentences into an SRS on the All Japanese All The Time blog.
Matt Colwell (Matt vs Japan) Creator

Systematised mining card formats and criteria in MIA and later in Refold's guides.
Migaku Company

Builds browser tooling whose core function is mining sentences from Netflix and YouTube into flashcards.
Tatsumoto Creator

Pseudonymous author of the current AJATT site, which publishes a detailed sentence-mining guide and card templates.

Learners who used this

Field notes whose stack included this approach. It is an account of what one person did, not evidence that it works — and note how few run a single method on its own.

A 1,480-day Anki streak ended, and the relief was the lesson · 48 months · blended with 2 other approaches

All field notes →

Related methods

Resources that use this

Anki

Multi-language

The gold-standard spaced-repetition flashcard app with community decks for every language.