Memorising the Qur’an Is Now More Structured with the Help of AI
JAKARTA — Memorising the Qur’an is both a spiritual endeavour and a complex cognitive process requiring consistency and strong memory retention. For centuries, memorisation methods at many Islamic educational institutions have generally relied on a one-size-fits-all approach.
Students are often given the same targets for new memorisation (ziyadah) and revision (muroja’ah), at the same pace and frequency, without accounting for differences in information-processing speed and individual memory capacity.
Artificial intelligence (AI) is opening up a new approach through the concept of a Personalised Learning Path (PLP). With this method, memorisation strategies can be designed automatically to better match each individual’s capacity.
Researchers Miftah Farid Adiwisastra, Irfan Darmawan and Dade Nurjanah, from Bina Sarana Informatika University and Telkom University, have pursued this opportunity by developing a structured dataset to model the learning process of Qur’an memorizers.
Gathering authentic data from Islamic boarding schools
AI systems need context-rich data to map learning habits accurately. The researchers therefore collected data from 350 students at three Islamic boarding schools in Indonesia between February and June 2025.
The data covered three main areas:
Demographic characteristics: students’ basic profiles, including age, sex and formal education level.
Current memorisation capacity: the cumulative number of juz memorised, the order of the surahs they could recall, years spent learning, and their target timeframe for completing all 30 juz.
Study habits and pace: the daily rate of adding new memorisation, the duration and frequency of muroja’ah, and the distribution of revision across four periods: morning, afternoon, evening and night.
Turning raw data into variables
Before computational analysis, field data were processed using scientific standardisation. The researchers applied Min-Max normalisation to bring all parameters onto a scale from 0 to 1 and prevent bias in variable weighting.
Memorisation achievement was also divided into three measured levels: 0–10 juz, 11–20 juz, and 21–30 juz. The researchers further used feature engineering to create new, context-specific metrics.
These metrics included an efficiency ratio, comparing the number of juz memorised with the number of years spent learning, and the daily frequency of revision (freq reviews).
Assessing memorisation status
The memorisation process involves changing internal conditions that cannot be measured by outward appearance alone. The researchers therefore selected the probabilistic Hidden Markov Model (HMM) algorithm to detect students’ latent cognitive states.
Memorisation status was divided into three levels: Beginner, Intermediate and Advanced.
Initial analysis found that 45 per cent of students were at Beginner level, 25 per cent were Intermediate and 30 per cent had reached Advanced. The transition-probability analysis also found that the Intermediate group had the most stable status, at 86.7 per cent, followed by the Advanced group at 73.1 per cent and the Beginner group at 56.7 per cent.
The figures provide empirical evidence that the transition to long-term memory requires gradual consolidation and cannot be accelerated instantly.
Model evaluation: balancing precision and caution
Tests of the HMM’s ability to detect students’ learning status produced varied results:
Precision: 71.46 per cent
Accuracy: 30.57 per cent
Recall: 30.57 per cent
F1 score: 37.70 per cent
Precision above 71 per cent indicates that each status recommendation generated by the system has a strong degree of reliability. However, the relatively low accuracy and recall point to a highly cautious or conservative algorithm.
The system tends to maintain predictions it considers safe until the data show a significant improvement in ability.
Towards adaptive Qur’an education
The availability of this structured data is an important step towards modernising Qur’an teaching methods. Initial testing suggests that combining the recommendation system with a web-based monitoring dashboard helps teachers draw up proportionate muroja’ah schedules and support students’ retention of memorised passages.
Teacher involvement through a feedback loop remains essential for validating the system’s output.
Future research priorities include integrating Automatic Speech Recognition technology tailored to tajwid pronunciation and adding more dynamic variables on daily routines. The collaboration between cognitive science, the pesantren tradition and AI offers the prospect of making the process of preserving the words of God more effective, measurable and suited to each learner’s capacity.
Source: jawawa.id (https://jawawa.id/newsitem/memorising-the-quran-is-now-more-structured-with-the-help-of-ai-1793004792)
Originally published by Republika (https://news.republika.co.id/berita/tni8gf416/menghafal-alquran-kini-lebih-terarah-berkat-bantuan-ai), translated from Indonesian