Sentence mining is a vocabulary method where you save whole sentences from real content (a show, a video, an article) that contain one unknown word or structure, then review them as spaced-repetition flashcards. Each card keeps its original context and, if you capture it, its audio, so you learn how the word is actually used, not just a dictionary gloss.
This guide covers how the method works step by step, which sentences are worth saving, what the research says and does not say about it, and how to fit it into a routine you can keep for months. It assumes you can already follow part of what you watch or read in your target language. If you cannot, the first step is easier content, not more cards.
Are you a Spanish speaker learning English? There is a separate Spanish version with examples built for you: Sentence mining: qué es y cómo hacerlo con series y videos.
Here is the method in four lines:
- Source: content you chose and mostly understand, not a textbook.
- Unit: one sentence with one unknown word or pattern.
- Card: the sentence, its meaning, and the audio or image from the scene.
- Review: spaced repetition, with a small and steady number of new cards per day.
What sentence mining is, and what it is not
The name describes the job. You dig through content you are already consuming and pull out the few sentences worth keeping, the way a miner picks ore out of rock. The rest of the video or article stays where it is and keeps working as input.
A mined sentence becomes a flashcard with three parts: the sentence, the thing you want to learn from it (one word, one phrase or one structure), and its meaning in that context. Good cards also carry the source: the audio of the line, a screenshot of the scene, or the page it came from.
Here is an example, written for this guide and not taken from a real show. You are learning Spanish, you are watching a series, and a character says a line you understand except for one piece:
| Field | Content |
|---|---|
| Sentence | Se me olvidó llamarte. |
| Target | se me olvidó + infinitive |
| Meaning in context | I forgot to call you (it slipped my mind; I did not do it on purpose) |
| Audio | The character's line, a few seconds long |
| Source | Series, episode, timestamp |
Every other word in that sentence is one you already know. That is what makes it a good card. The unknown part is small, so the meaning is easy to work out, and the card is quick to review.
Sentence mining is not:
- A word list. Saving every unknown word with a dictionary gloss builds a pile of isolated facts.
- A replacement for input. The cards are a safety net under the watching and reading, not the main activity.
- Tied to one app. Any tool that schedules reviews works. Anki is the best known, but a spaced-repetition system inside another app does the same job.
Why mine sentences instead of single words
A single word with a translation tells you what a word can mean. A sentence tells you how the word behaves: which words it sits next to, which grammar it triggers, how formal it sounds and how it is pronounced in natural speech. The Spanish example above shows it. A dictionary gives "olvidar" as "to forget". The sentence adds something the entry does not: Spanish also says it as something that happens to you, with "se me olvidó".
There is research behind the idea that repeated encounters with a word in context build knowledge of it. In a 2007 study in Applied Linguistics, Stuart Webb had 121 Japanese learners of English meet target words in reading texts 1, 3, 7 or 10 times, then tested several aspects of word knowledge, including spelling, meaning and form, grammatical function, syntax and association. Each increase in the number of encounters produced larger gains in at least one aspect, ten encounters could produce sizeable gains, and full knowledge of a word may need more than ten.
Two cautions apply. The study used made-up words inside reading texts, so it measures the effect of encounters, not of flashcards. And it does not show that sentence cards beat word cards. What it supports is narrower, and it is useful for mining: a word needs many meetings, and each meeting can add information about a different aspect of how the word works. Spaced reviews of a sentence card are a cheap way to get several of those meetings. The real content you keep watching supplies the rest, in new contexts.
Audio matters for the same reason. A written sentence does not tell you where the stress falls or how words blur together in fast speech. A card with the original line lets you hear it every time it comes up.
The 1T rule: one unknown per sentence
Learners in immersion communities call a sentence with exactly one unknown element a 1T sentence, for "one target". Refold's sentence mining page describes the rule like this: mine sentences with one unknown word, or possibly two, that look useful or interesting, and skip sentences with three or more, because you do not understand them well enough yet to learn from them. It also advises skipping very niche words that are unlikely to come back.
The logic is practical. With one unknown, the rest of the sentence is a scaffold. You can often guess the meaning from context and confirm it with a lookup. With three unknowns, you are memorizing a puzzle, and the card will be hard to answer for weeks.
The rule borrows its vocabulary from Stephen Krashen's i+1, but the two are not the same claim. In Principles and Practice in Second Language Acquisition (1982), Krashen states that moving from stage i to stage i+1 requires understanding input that contains i+1, with the help of context and extra-linguistic information. He adds that the input need not contain only i+1: if the input is understood and there is enough of it, i+1 is provided automatically, so deliberately aiming at it is not necessary. Mining 1T sentences is a card-selection habit. It is not a test of Krashen's theory.
What you can take from Krashen is the part that fits mining: understand the message first, and get a lot of it. Cards for unknown words speed up what the content is already doing. If you catch yourself mining before you understand what is happening in the scene, the content is too hard for now.
The sentence mining workflow, step by step
- Pick content you mostly understand. A series you can follow with subtitles, a video where you catch the gist, an article you can read without translating every line. If nearly every sentence has several unknown words, switch to something easier before you mine anything.
- Consume first, mine second. Watch or read in chunks without stopping at every unknown word. Mark a sentence only when it is useful and you want to keep it.
- Choose 1T sentences. One unknown word or structure, everything else familiar. Skip names, one-off technical terms and lines where you understand almost nothing.
- Look up the meaning in context. The first dictionary entry is often the wrong sense. Check what the word means in this sentence. Our guide to contextual versus generic dictionary lookups shows why this matters.
- Save the card. Sentence, target, meaning, and the source media (audio, image or both). Keep it short enough to review quickly.
- Review on schedule. The scheduler decides when each card comes back. Your job is to show up and answer honestly.
- Go back to the content. The point is to understand the next episode better than the last one. If reviews start eating the time you spend on input, cut new cards.
What a good sentence card contains
A card is a small container for one lesson. These are the fields that earn their place:
| Field | What to put | Why |
|---|---|---|
| Sentence | The full line, trimmed to the clause that carries the meaning | Shows grammar and word combinations in use |
| Target | One word, phrase or structure, marked or highlighted | Keeps the card to one question |
| Meaning | The sense in this sentence, in a few words | Prevents learning the wrong sense |
| Audio | The original line | Trains your ear on natural speech |
| Image | A frame from the scene or the page | Gives a second memory cue |
| Source | Title and timestamp, or the URL | Lets you return to the scene if the card confuses you |
On card format there are two common choices. A recognition card shows the sentence and asks for the meaning. A cloze card hides the target and asks you to recall it. Piotr Wozniak, the researcher behind SuperMemo's scheduling algorithms, argues in his twenty rules of formulating knowledge (1999) for the minimum information principle: make each item as simple as possible, and split complex items into smaller ones. He also calls cloze deletion easy and effective.
For mining, that translates into one target per card. If a sentence holds two things worth learning, make two cards from it instead of one card with two answers. A reasonable order is to start with recognition cards, which are easier to answer and help you keep the habit, and add cloze cards for words you keep recognizing but cannot produce yet.
What to mine and what to skip
| Mine | Skip |
|---|---|
| Words you met more than once, in different scenes | Character names and place names |
| Verbs together with the preposition or pattern they take | Technical terms from a topic you will not meet again |
| Fixed expressions, saved as a single unit | Sentences with three or more unknown words |
| Structures that keep confusing you, like "se me olvidó" | Words you realize you already know |
| Words that match your goals and the topics you want to read or talk about | Lines you cannot make sense of even after a lookup |
The same Refold page advises deleting cards for words you turn out to know already, and not letting unreviewed cards pile up. Both habits keep the queue made of things you still need.
Learn Languages with YouTube
Watch videos in your target language, auto-capture words into flashcards, and track your progress — all in one place.
Start Free →Mining from video, articles and audio
Video is a convenient source because it gives you everything a good card needs: a sentence, native audio and a visual cue. Prefer subtitles uploaded by the creator over automatic ones when you have the choice. YouTube Help states that automatic captions are machine-generated and can misrepresent the spoken content because of accents, mispronunciations or background noise. A card built on a wrong caption teaches the wrong sentence.
Articles and books give you time. You can read at your own speed and the text does not disappear, so mining is less disruptive. You lose the audio, so use a text-to-speech or the source's own audio if your tool offers it, and always save the page or book reference.
Podcasts and audio-only content are the hardest to mine, because without a transcript you cannot read the line you want to keep. If a transcript exists, use it. If not, treat audio as pure listening practice and mine from material that has text.
Why the cards need spaced repetition
Saving a sentence does nothing for memory by itself. Review does. Spaced repetition means reviewing each card at increasing gaps instead of all at once. Spacing is one of the best-studied effects in memory research. Cepeda and colleagues (2006) reviewed 839 assessments of distributed practice from 317 experiments in 184 articles. Among their findings, in verbal recall tasks: the gap between study sessions that gives the best retention gets longer as the time you need to remember gets longer.
In practice, if you want to remember a word next month, a review spread over days and weeks beats ten repetitions in one evening. Scheduling software does the arithmetic for you.
The Anki manual explains the two ideas behind this. Active recall means you try to retrieve the answer instead of rereading it, which strengthens memory and shows you what needs more review. Spacing means that when you recall a card successfully, the interval before the next review gets longer. Anki's classic scheduler is based on SM-2, and since version 23.10 it also offers FSRS as an alternative, where the main setting is the desired retention (90% by default). Metheus schedules reviews with FSRS as well.
How many new cards per day? There is no research-based number, and any figure is a rule of thumb, including the one below. What is certain is the arithmetic: every new card creates reviews for months. Refold warns against mining 30 or more cards in one session. A safe way to start is 5 to 10 new cards per day, then adjust after two weeks according to how long your daily reviews take. If the queue already feels heavy, our guides on fixing Anki overload and making reviews less boring cover pruning, limits and session design.
Manual, semi-automatic and integrated mining
The method is the same however you capture cards. What changes is how much work each card costs, and that cost decides whether you keep going.
- Manual. You pause the video, copy the subtitle, look up the word, take a screenshot, trim the audio and paste everything into your flashcard app. It works and gives you full control. It also interrupts what you are watching on every card.
- Semi-automatic. Subtitle players and dictionary extensions send cards to Anki through add-ons such as AnkiConnect. This removes the copy-paste but adds a chain of tools to maintain, and when one link breaks, media goes missing. Our guide to exporting Anki decks with audio and images covers the usual failures.
- Integrated. Lookup, capture and review live in one system, so one action saves the card and the scheduler picks it up.
Do not take our word for the difference. Time yourself. Mine ten cards the way you do it now and write down the total minutes. Then repeat with another setup and compare. The setup with the lower cost per card is the one you will still be using in three months.
If mining makes watching feel like homework, you are probably paying too much per card. Our guide to avoiding sentence mining burnout goes through the causes and the fixes.
Sentence mining with Metheus, step by step
Metheus is built around this loop. This is how it works today:
- Install the Metheus browser extension. It is free and available for Chrome, Firefox and Edge.
- Open a video on YouTube, Netflix, Amazon Prime Video, Disney+ or Viki with subtitles in your target language. The subtitles become interactive: you can click a word to see its meaning in context, and words are highlighted according to your vocabulary status. On other pages, turn the extension on from its popup.
- When a line has one unknown word, mine it. The card can store the sentence, a translation, a screenshot of the scene and the audio of the line.
- The card goes to your library in the Metheus web app, where it joins the reviews scheduled with FSRS.
- Keep watching. Review in the web app when it suits you.
All of this is in the free plan: immersion, mining, FSRS reviews, your own decks and the full extension. Premium adds the later course modules and is not needed for mining.
A few limits are worth knowing. Interactive subtitles depend on the platform serving a subtitle track in your target language, and some DRM-protected content is restricted. Metheus is a browser tool, so there is no desktop app. If you already keep a deck in Anki, the web app has an import page that accepts .apkg files.
Common mistakes
- Mining every unknown word. This produces a long, low-value queue. Mine what is useful and likely to return.
- Mining sentences with several unknowns. The card becomes a puzzle you will fail again and again.
- Saving cards and skipping reviews. A card that is never reviewed teaches nothing. If you cannot review for a few days, pause new cards.
- Pausing at every line. Constant stopping breaks comprehension. Mark the sentence and keep going, or use a tool that captures without stopping the video.
- Cards with no source. Without the audio or a pointer to the scene, you lose half the context that made the sentence worth saving.
- Overlong sentences. If a line is a paragraph, trim it to the clause that holds the target.
- Treating cards as the main activity. Reviews support input. If reviews take more time than watching, cut new cards.
What sentence mining does not do
Sentence mining builds recognition of vocabulary and patterns. By itself it does not train you to speak or to keep up with fast conversation; those skills need their own practice. It does not teach grammar systematically either, although 1T cards on structures, like the Spanish example earlier, help with patterns you keep meeting.
The evidence is also indirect. We could not find a controlled study that tests sentence mining as a complete method. The case for it rests on three pieces that do have support: understandable input (Krashen), repeated encounters with words in context (Webb) and spaced review (Cepeda and colleagues). That is a reasonable foundation. It is not proof that this exact routine beats every other.
It is also optional. Some learners progress well with extensive reading and listening and very few cards. If you do not enjoy reviewing, make the review share small instead of dropping the input.
A four-week plan to start
- Week 1: watch and mark. Spend 20 to 30 minutes a day on content you mostly understand. Mine up to 5 sentences a day and notice how many lines are 1T.
- Week 2: add rhythm. Keep to 5 new cards a day and do all your reviews. Time your review session.
- Week 3: adjust. If reviews take more than about 20 minutes, drop to 3 new cards a day for a week. If they take under 10 and you want more, move up toward 10. Delete cards that bore you or that you already know.
- Week 4: check the result. Watch an episode or read a text from the same source. How many of the mined words do you recognize without a lookup? If few, review your 1T discipline and your card quality before adding more.
These numbers are starting points, not research results. Change them to fit your time.
Sources and further reading
- Krashen, S. (1982). Principles and Practice in Second Language Acquisition. The original statement of the input hypothesis and i+1, free on the author's site.
- Webb, S. (2007). The effects of repetition on vocabulary knowledge. Applied Linguistics, 28(1), 46-65.
- Cepeda, N. J., Pashler, H., Vul, E., Wixted, J. T., and Rohrer, D. (2006). Distributed practice in verbal recall tasks: A review and quantitative synthesis. Psychological Bulletin, 132(3), 354-380.
- Wozniak, P. (1999). Twenty rules of formulating knowledge. SuperMemo.
- Anki manual: Background and Deck options (FSRS).
- Refold: Sentence mining, a community roadmap page and the source of the 1T description above.
Start with one video
Pick one short video you can mostly follow, find one sentence with one unknown word, and turn it into a card. That is sentence mining. Everything else is repetition and tuning.
Create your free Metheus account and mine your first sentence from the video you were going to watch anyway.
FAQ
What is sentence mining?
Sentence mining is a vocabulary method where you save whole sentences from real content, such as a show, a video or an article, that contain one unknown word or structure, and review them as spaced-repetition flashcards. Each card keeps the original context, and often the audio, so you learn how the word is used.
What is a 1T sentence?
A 1T sentence is a sentence with exactly one unknown element, the "one target". The rest of the sentence is familiar, so you can work out the meaning from context and the card is quick to review. Refold suggests mining sentences with one unknown word, or possibly two, and skipping those with three or more.
How many sentences should I mine per day?
There is no research-based number. A common starting point is 5 to 10 new cards per day, adjusted after two weeks according to how long your reviews take. Refold advises against mining 30 or more cards in one session because unreviewed cards pile up.
Do I need Anki to do sentence mining?
No. Sentence mining only needs a way to save sentences and a spaced-repetition scheduler. Anki is the best-known option. Metheus also schedules reviews with FSRS inside its web app, and its import page accepts Anki .apkg files if you already have a deck.
Is sentence mining good for beginners?
It works once you can follow part of the content, because you need sentences where only one piece is unknown. If almost every line has several unknown words, start with easier material or core vocabulary first, then begin mining.
Which sites does Metheus support for sentence mining?
The Metheus extension is free for Chrome, Firefox and Edge, with interactive subtitles and mining on YouTube, Netflix, Amazon Prime Video, Disney+ and Viki. On other pages you can turn the extension on from its popup. Availability depends on the platform offering a subtitle track in your target language.
2026 Immersion Report
Language learning in 2026 isn't about finishing daily streaks. The fastest path to fluency combines three elements: comprehensible input from real media you enjoy, active vocabulary capture while consuming that content, and spaced repetition to lock words into long-term memory. Gamified apps build the habit. Immersion tools build the fluency. The most effective learners use both—in that order.
Read the Full 2026 Comparison GuideStart with Metheus
Watch videos in your target language, auto-capture words into flashcards, and track your progress — all in one place.
Get Started Free →
