PERANCANGAN SISTEM PENGENALAN TULISAN TANGAN PADA JAWABAN ESAI MENGGUNAKAN METODE CNN-LSTM BERBASIS ANDROID

Elsa Apriani; Nunik Pratiwi;

Perancangan Sistem Pengenalan Tulisan Tangan pada Jawaban Esai
Menggunakan Metode CNN-LSTM Berbasis Android

Elsa Apriani

Perkembangan teknologi mendorong inovasi pendidikan, salah satunya sistem pengenalan tulisan
tangan untuk mempercepat penilaian jawaban esai. Penelitian ini merancang aplikasi Android yang
mengenali tulisan tangan siswa menggunakan metode Convolutional Neural Network (CNN) dan
Long Short-Term Memory (LSTM). Sistem dikembangkan dengan pendekatan prototyping melalui
tahapan identifikasi kebutuhan, perancangan antarmuka, implementasi fitur, dan pengujian. Hasil
evaluasi menunjukkan akurasi 60,37%, Character Error Rate (CER) 16,84%, dan Word Error Rate
(WER) 78,41%. Meskipun WER masih tinggi, akurasi karakter cukup baik untuk tahap awal
pengembangan dan memberikan dasar yang menjanjikan untuk perbaikan sistem ke depan.
Pengujian menggunakan Black Box Testing memastikan semua fitur berjalan sesuai fungsinya.
Sistem ini diharapkan mempermudah pengajar dalam mengoreksi jawaban esai secara lebih efisien,
cepat, dan konsisten, serta mendukung digitalisasi penilaian di lingkungan pendidikan.
Kata kunci: Android, CNN, LSTM, OCR, Tulisan Tangan

Design of Handwriting Recognition System for Essay Answers Using CNN-
LSTM Method Based on Android

Elsa Apriani

Technological developments drive educational innovation, one of which is a handwriting
recognition system to accelerate essay answer assessment. This study designs an Android
application that recognizes students' handwriting using the Convolutional Neural Network (CNN)
and Long Short-Term Memory (LSTM) methods. The system was developed using a prototyping
approach through the stages of identifying needs, designing interfaces, implementing features, and
testing. The evaluation results showed an accuracy of 60.37%, Character Error Rate (CER) of
16.84%, and Word Error Rate (WER) of 78.41%. Although the WER is still high, character accuracy
is good enough for the early stages of development and provides a promising basis for future system
improvements. Testing using Black Box Testing ensures that all features run according to their
functions. This system is expected to make it easier for teachers to correct essay answers more
efficiently, quickly, and consistently, as well as support the digitalization of assessment in the
educational environment.
Keywords: Android, CNN, LSTM, OCR, Handwriting