Sri Masyitah, Rizka Sari, Abdur Rahman Purba, Muhammad Ilham, Fitri Mawaddah Bako

ANALISIS SENTIMEN ULASAN SISWA TERHADAP PEMBELAJARAN BAHASA ARAB MENGGUNAKAN DEEP LEARNING DI SDIT KHAIRUR RAHMAN

Introduction

Analisis sentimen ulasan siswa terhadap pembelajaran bahasa arab menggunakan deep learning di sdit khairur rahman. Analisis sentimen ulasan siswa SDIT Khairur Rahman tentang pembelajaran Bahasa Arab daring dengan Deep Learning (LSTM). Temukan 52% sentimen negatif, sarankan perbaikan interaksi & media. Akurasi model 88.3%.

Abstract

ABSTRAK: Deep learning mendorong lembaga pendidikan untuk mengalihkan pembelajaran dari tatap muka ke daring, yang menghasilkan respons beragam dari siswa dalam hal motivasi, efektivitas, dan interaksi. Meskipun penelitian sebelumnya telah mengkaji efektivitas pembelajaran daring secara umum, hanya sedikit yang meneliti sentimen siswa menggunakan analisis teks berbasis deep learning. Penelitian ini bertujuan untuk menganalisis ulasan siswa terhadap pembelajaran daring di SDIT Khairur Rahman melalui klasifikasi sentimen menggunakan model Long Short-Term Memory (LSTM). Sebanyak 20 ulasan siswa dikumpulkan dan diberi label secara manual sebagai positif, netral, atau negatif. Data kemudian melalui proses prapengolahan teks, tokenisasi, dan pelatihan menggunakan algoritma LSTM untuk mengklasifikasikan sentimen secara otomatis. Model ini mencapai akurasi sebesar 88,3%, dengan nilai precision, recall, dan F1-score masing-masing sebesar 87,1%, 86,5%, dan 86,8%. Analisis menunjukkan bahwa 52% ulasan siswa mengungkapkan sentimen negatif, 28% positif, dan 20% netral. Temuan ini menegaskan perlunya peningkatan interaksi guru dan siswa, kualitas media pembelajaran, serta dukungan teknis dalam sistem pembelajaran daring, sehingga memberikan wawasan berharga bagi pendidik dan pembuat kebijakan.

Review

This paper presents an analysis of student sentiment towards online Arabic language learning at SDIT Khairur Rahman, utilizing a deep learning approach with Long Short-Term Memory (LSTM) models. The study addresses a pertinent gap by focusing on student sentiment in a specific educational context, moving beyond general effectiveness studies of online learning. The application of deep learning for text-based sentiment analysis in this domain is a noteworthy endeavor, aiming to automatically classify reviews as positive, neutral, or negative. The reported accuracy of 88.3% for the LSTM model, alongside detailed precision, recall, and F1-score metrics, suggests a seemingly effective classification performance, leading to practical recommendations for improving teacher-student interaction, media quality, and technical support. However, a critical limitation significantly impacts the robustness and generalizability of the findings: the dataset consists of merely 20 student reviews. This exceedingly small sample size is highly problematic for training deep learning models, which typically require vast amounts of data to learn complex patterns and avoid overfitting. Consequently, the reported high accuracy rates may not truly reflect the model's performance on unseen data, and the classification of sentiments (52% negative, 28% positive, 20% neutral) is based on an insufficient number of data points to draw reliable conclusions or suggest broad policy implications. The justification for employing a sophisticated deep learning algorithm like LSTM with such a limited dataset is weak; simpler machine learning or even rule-based methods might have been more appropriate or at least explicitly compared. Despite the methodological concerns regarding sample size, the study does offer initial insights into the perceptions of students at the specific institution. For future research, it is imperative to significantly expand the dataset of student reviews to validate the deep learning model's efficacy and ensure the findings are representative and robust. Further work could also explore comparative analyses with traditional machine learning algorithms to determine the optimal approach for this specific application, given varying data availability. This foundational effort provides a valuable starting point for SDIT Khairur Rahman to understand its students' experiences, but substantial improvements in data collection and methodological rigor are necessary to elevate the study's impact and generalizability.

Full Text

Comments

You need to be logged in to post a comment.