Platform Adaptive Learning Berbasis NLP untuk Pembelajaran Literasi Siswa Disleksia Sekolah Dasar

Abstract

Penelitian ini bertujuan mengembangkan platform pembelajaran berbasis web yang mengintegrasikan Adaptive Learning dan Natural Language Processing (NLP) untuk mendukung pembelajaran membaca dan menulis siswa sekolah dasar dengan gangguan disleksia. Permasalahan utama yang melatarbelakangi penelitian ini adalah terbatasnya platform pembelajaran digital yang mampu menyesuaikan materi, latihan, dan rekomendasi pembelajaran berdasarkan respons teks siswa disleksia. Penelitian ini menggunakan metode pengembangan perangkat lunak model Waterfall yang meliputi analisis kebutuhan, perancangan sistem, implementasi, pengujian, dan pemeliharaan. Sistem dikembangkan dengan antarmuka web AdaptLearn, backend berbasis Gin Framework, frontend berbasis Next.js, basis data PostgreSQL, cache Redis, serta keamanan autentikasi menggunakan bcrypt. Modul NLP digunakan untuk menganalisis jawaban teks siswa dan mengklasifikasikan tingkat pemahaman ke dalam kategori Paham, Cukup Paham, dan Belum Paham. Hasil pengujian menunjukkan bahwa seluruh fitur utama sistem berjalan sesuai skenario black-box testing. Evaluasi model NLP menghasilkan accuracy sebesar 0,99 dengan nilai precision, recall, dan F1-score yang tinggi pada seluruh kategori. Hasil usability testing menggunakan System Usability Scale memperoleh skor rata-rata 73,33 yang termasuk kategori baik dan acceptable. Dengan demikian, platform yang dikembangkan layak digunakan sebagai alternatif media pembelajaran literasi dasar yang lebih personal, adaptif, dan inklusif bagi siswa disleksia.

This study aimed to develop a web-based learning platform integrating Adaptive Learning and Natural Language Processing (NLP) to support reading and writing instruction for elementary school students with dyslexia. The study was motivated by the limited availability of digital learning platforms that can adjust learning materials, exercises, and recommendations based on dyslexic students’ text responses. This research used the Waterfall software development model, consisting of requirements analysis, system design, implementation, testing, and maintenance. The system was implemented as the AdaptLearn web platform using a Next.js frontend, Gin Framework backend, PostgreSQL database, Redis cache, and bcrypt-based authentication security. The NLP module was designed to analyze students’ written responses and classify their understanding into three categories: Understanding, Partially Understanding, and Not Yet Understanding. The Adaptive Learning mechanism then used the NLP results and learning interactions to recommend remedial or advanced learning materials. The results showed that all major system functions performed as expected in black-box testing. NLP model evaluation produced an accuracy of 0.99, with high precision, recall, and F1-score across all categories. Usability testing using the System Usability Scale obtained an average score of 73.33, indicating good and acceptable usability. Therefore, the developed platform is feasible as an alternative digital learning medium that supports more personalized, adaptive, and inclusive literacy learning for students with dyslexia.