gap_fill
Fill the gap in a sentence
Your term is taken out of a complete sentence, and you put it back. It checks whether you can use the word in a sentence, not just remember what it means.
MemorizeKit Practice
AI writes questions from your own terms.
Practice takes the vocabulary, phrasal verbs, and structures you've saved and has AI turn them into contextual questions — not the exact line you wrote down. You answer, your answer is graded, and each finished term sends exactly one result to your review schedule.
AI does three jobs, and only three.
Practice doesn't let AI run the whole session. AI is called in three specific places, and each output is checked before it reaches you.
Write
Put the term in a new context
AI writes questions from the term's data — text, meaning, part of speech — instead of repeating your notes word for word. Generated questions are saved and never change.
Grade
Understand what you wrote, not just match strings
Typed translations are often right in meaning but different in wording. When the rules can't decide, AI reads your sentence and returns a structured verdict.
Prepare
Refill the question bank behind the scenes
After a run uses up ready questions, MemorizeKit orders more in the background so you don't wait next time. If that fails, the run you just finished isn't affected.
You never see the AI provider's name or the grading prompt, and you can't edit generated questions. Practice only returns questions and results.
One Practice run, five steps.
You press Practice once. Behind it, these five steps run in order, and each one starts only when the previous one has fully finished.
Step 1
Choose terms to ask
From My Library, Practice takes terms that are due first, then terms you haven't studied. From a Lesson, the scope narrows to that Lesson — or just its starred terms.
Step 2
Use ready questions, have AI write the rest
Every term has a question bank. If it's full, your run starts right away. If not, MemorizeKit has AI write more and saves them before returning the run — you never get a half-built run.
Step 3
Spread questions evenly across terms
Your question count is spread in rounds, so a term can appear more than once when needed. Ten questions over three terms split 4 / 3 / 3 instead of piling onto the first.
Step 4
You answer, the system grades
Multiple choice and gap-fill are graded by rules, instantly. Only typed translations the rules can't decide go to the AI evaluator. You see the result with a reference answer.
Step 5
A finished term goes to your review schedule
When a term's last question is graded, Practice combines that term's questions into exactly one result for the memory layer. A half-finished term sends nothing.
Two question families, two ways to answer.
For now, Practice asks in two families. Each can appear as multiple choice or typed, and can ask forward or in reverse.
translation
Match words and meanings, both ways
Forward direction gives you the word and asks for the meaning. Reverse direction gives you the meaning and asks for the word — the harder way, because there's nothing to rule out.
- Multiple choice
- Four numbered choices, picked by mouse or keyboard. Check only turns on once you've chosen.
- Typed answer
- You write the answer yourself. This is where the AI evaluator earns its keep, when your sentence is right in meaning but worded differently.
The family list is open. A new family that reuses these two answer modes needs no change to the Practice screen or the review schedule.
Rules grade first; AI steps in only when the rules can't.
Multiple choice, exact gap-fills, and Don't know all have a clear verdict, so MemorizeKit grades them itself — fast, cheap, and the same every time. The AI evaluator only gets what's left.
The answer is never sent to your device before you respond. Only afterwards does Practice return the result with a reference answer.
Correct
Your answer matches the term's recorded meaning. Practice says whether the verdict came from Automatic evaluation or AI evaluation.
Keep practicing
Your answer is wrong. Practice puts it next to the reference answer so you can see the difference.
Don’t know
You choose to skip. No need to guess — but the question still counts toward the term's overall result.
Item complete
A term's last question was just graded. Practice shows the Item result — Correct or Wrong — before moving on.
You pick the scope; Practice picks the order.
There are three ways to open Practice. None of them make you choose terms one by one.
- From My Library
- All your terms. Due terms are asked first, then terms you haven't studied.
- From a Lesson
- The same order, but only inside the Lesson you have open. You can only practice Lessons you own.
- Starred in a Lesson
- Only terms you starred, including unstudied ones. Practice never tops up with unstarred terms.
5 to 25 questions per run
The number you set is a question count, not a term count — 20 by default. Practice returns exactly that many distinct questions, or says plainly that there's nothing to practice yet.
Spaced repetition
Each term sends exactly one result.
A term can be asked three times in one run. Those three questions don't become three review entries — Practice combines them into one verdict for the term, then sends it.
- Every question of a term correct, and the term counts as correct.
- One wrong question makes the term wrong — no partial credit.
- A term with unanswered questions sends no result.
Practice doesn't calculate review dates. It sends a verified result, and MemorizeKit's memory layer keeps the schedule — on the same Level ladder Learn uses.
Practice runs aren't kept.
Practice forgets on purpose. What's kept is the shared question bank and each term's result — not a log of every answer you typed.
- Your answers aren't stored
- What you type is used for grading, then discarded. There's no answer history and no separate Practice skill profile.
- Refreshing starts a new run
- Progress lives in your browser. Refreshing gives you a different run, and can't undo results already recorded in your schedule.
- Old questions retire themselves
- Edit the source term and questions built from the old version stop being used. You start again with questions that match the current term.
When AI isn't ready, Practice says so.
AI services are sometimes busy or failing. Practice doesn't paper over the gap with a half-run or a guessed result.
Nothing to practice
There's no eligible term in this scope. Practice stops there and takes you back to the Lesson — no AI call, nothing created.
Not enough questions yet
Even one missing question keeps the run closed. You get a retryable error instead of a shorter run than you asked for.
This answer can't be graded yet
The AI evaluator didn't return a valid verdict. The answer you just sent can be reused, so retrying doesn't lose progress.
The source term just changed
You edited the term mid-run. Practice refuses to grade against the old version and invites you to start a new run.
Learn to memorize. Practice to use it.
Learn asks about the exact terms you saved, to fix them in memory. Practice puts those terms into new sentences written by AI, to see whether you can really use them. Both share one Level ladder.
Watch AI write questions from one of your terms.
Open My Library or a Lesson and start Practice to see contextual questions, grading, and each term's result.
Open My Library
No account yet? Sign up free