How to Choose Sentences for Sentence Mining

The best sentence to mine is not the sentence with the most unknown language. It is the sentence that gives one useful target enough natural context to become a fair, repeatable review.

The five-part selection filter

1. Useful

Would you value understanding or using this expression again? Prefer language that appears in situations, genres, or relationships relevant to you. Frequency lists can guide attention, but personal usefulness determines whether the card will still feel worthwhile after many reviews.

2. Mostly understandable

Can you explain the sentence after checking one main uncertainty? A card that needs five definitions, a paragraph of cultural context, and a grammar lecture is a research project.

Aim for one main target rather than a fixed unknown-word count. Proper names, transparent cognates, and predictable inflections should not be treated the same as several unrelated new concepts.

3. Focused

Name one reason to save the sentence:

• the phrase resultar ser;
• a past-tense contrast;
• reduced pronunciation;
• a particle; or
• a useful response pattern.

If three reasons compete, choose the strongest one. You can preserve other observations in Notes without creating sibling cards.

4. Verifiable

Can you trust the sentence text, audio, meaning, and source? A clean timestamp and a short audio boundary make later correction possible. Skip a noisy or uncertain line when the uncertainty is exactly where the learning target lives.

5. Worth reviewing

Imagine the card appearing in six months. Would you be glad to see it, indifferent, or annoyed? That forecast is imperfect, but it prevents “interesting for five seconds” from becoming scheduled work.

Once a candidate becomes a card, use the daily sentence-card review workflow to grade it consistently and repair prompts that cannot be judged fairly.

Strong and weak candidates

Strong: one contextual pattern

Llevo dos años estudiando español.

The sentence is short, natural, and gives a clear example of llevar + time + gerund. A recognition card can show the sentence and ask for contextual understanding.

Weak: a paragraph with four highlights

A long monologue may contain several useful items, but one card cannot diagnose which one failed. Choose the best complete clause or wait for a clearer example.

Strong: audio reveals a gap

You understand the subtitle but repeatedly miss the connected speech. Save an audio-first card, keep the transcript on the back, and mark the exact sound contrast as the focus.

Weak: the image gives away the answer

A sentence about pouring rain paired with an obvious rain image may test picture recognition. Put that image on the back unless the scene is necessary to identify a referent.

Strong: the scene bounds production

Instead of “Translate: I forgot,” use a specific situation and name the construction you want to practice. A fair production prompt narrows reasonable answers without spelling them out.

Choose from video without breaking flow

Use a two-pass system:

1. During viewing, save only a timestamp or lightweight marker.
2. After a natural stopping point, revisit the candidates.
3. Apply the five-part filter.
4. Keep the best few, not every marked line.
5. Trim audio and capture context only for accepted sentences.

Set a capture budget before watching. If five cards per episode already fill your available review capacity, the sixth candidate must replace a weaker one rather than silently expanding the queue. The card-budget guide turns observed review capacity into a practical limit.

For YouTube material you are permitted to use, captions help locate candidates but must be checked against the sound. The subtitle-to-Anki workflow covers cue selection and timing.

Ten-second decision checklist

Accept the sentence when you can answer yes to at least the first four:

• I know the one main target.
• I understand the rest well enough to learn from it.
• The sentence sounds natural in this context.
• I can verify the text, meaning, and source.
• The audio boundary is clean if sound matters.
• The card tests a skill not already covered.
• I would value seeing it again next month.

Skip or postpone when:

• several unrelated unknowns compete;
• the transcript is uncertain at the target;
• the context makes several answers equally valid;
• the card exists only because capture was easy; or
• your current review budget is already full.

Selection is not loss. The video, book, or conversation remains available as language exposure. Anki should receive only the moments that benefit from scheduled retrieval.

Sentence selection FAQ

What is a good sentence to mine?

One that is mostly understandable, contains a useful main target, provides enough context to verify meaning, and deserves future reviews.

How many unknown words should it have?

Use no rigid number. Prefer one main uncertainty; skip a sentence when several unknowns require a long explanation.

Should mined sentences be short?

They should be focused, not mechanically short. Keep enough language for natural meaning and remove unrelated dialogue or clauses.