Checking hifz with your phone: a practical memorization test for parents and teachers
Every hifz teacher and every parent of a hafiz-in-training knows the drill: pick an ayah somewhere in the juz, read the first words, and ask the student to carry on. It is the best test of memorization there is, and it needs an adult who knows the passage well enough to hear a slip. The Hifz test in recite.nabilmh.com does the asking and does a first check, so the adult can listen instead of following along in a mushaf.
How the test works
- Choose a range: one or several juz, a span like 67:1 to 67:30, or a set of pages, and how many ayahs the student should recite after the question (1 to 3). The choice is remembered for next time.
- The app shows a random ayah from the range, large, with its surah and number. It does not repeat an ayah until the whole range has been used, and it never asks the last ayah of a surah (that one is the answer to the ayah before it).
- The student continues with the next ayah from memory.
- The adult marks it: Correct, Needs work or Skip. They can reveal the next ayah, then the one after, to check.
The score runs in the header, missed ayahs are listed at the end of the round and can be retried, and a short history of rounds (date, range, score) stays on the device.
Listen: the automatic check
Press Listen and the app checks the continuation itself. It starts the recitation tracker at the first word of the ayah the student should recite, keeps the expected words hidden, and shows each word only as the student reaches it. At the end it says either "followed to the end" or "check this", with the reason:
- the student went to another ayah ("it sounded like 67:4"), a typical mutashabihat slip;
- they skipped over expected words;
- the tracker lost the place;
- they stopped for 6 seconds while still speaking, or listening stopped early.
The expected word where it went wrong is shown in red. It is a helper; the adult still decides with Correct or Needs work.
How reliable is the check?
We test it on recordings with known timings: an ayah recited right after the question (should pass), the same audio when a different ayah was expected (should be flagged), the right ayah stopped halfway (should be flagged), and a skipped ayah (should be flagged). On our current test set:
| Scenario | Handled correctly |
|---|---|
| right continuation, should pass | 56 of 61 |
| wrong ayah, should be flagged | 60 of 60 |
| stopped halfway, should be flagged | 61 of 61 |
| skipped an ayah, should be flagged | 58 of 58 |
The misses are on the side that matters least: a correct continuation sometimes gets "check this", and the adult confirms it was fine. A wrong continuation getting a pass is what would be harmful, and on this set it did not happen. These are our own test recordings; TODO: repeat the measurement on children's voices specifically, which are harder for every speech model.
Why a phone, why on the device
A hifz class has many students and one teacher. A phone running the test lets students revise in pairs or at home, and the teacher can spend their attention on the students who get "check this". Because the speech model runs on the phone (how), a child's recitation is never uploaded anywhere.
Mistake history
When the check flags a word, the mistake (the ayah, the word, expected versus heard) goes into a Mistake History on the device, so patterns show up over weeks: the same ayah confused every time, the same word skipped. With tajweed correction (coming soon) the same history also collects tajweed findings such as a missing shadda.
For developers and schools
The Hifz test is part of the Naml tracker code. If you run a hifz programme and want it inside your own app, with your own ranges and reports, tell us what you need.