Pengembangan Dataset Berbahasa Indonesia untuk Automated Essay Scoring (AES) dengan Fitur Feedback melalui Implementasi Teknik Fine-Tuning BERT pada Platform Website

Abstract

Natural Language Processing (NLP) is a branch of artificial intelligence (AI) that facilitates how computers can interact with humans using language that we understand, either in text or voice. One of the applications of NLP is Automated Essay Scoring (AES), which aims to provide automatic scoring of essays with the right accuracy. This research aims to build a Transformer-based AES model, specifically using the IndoBERT model specifically designed for Indonesian. The training dataset is manually created to ensure accurate and relevant data, while avoiding potential meaning errors that often occur due to the translation process from foreign languages. The model is designed to be able to generate scoring based on several scoring rubrics, namely the relevance of the answer to the prompt, correlation between sentences, essay length, and vocabulary richness. The last stage is evaluation, utilizing the Mean Squared Error (MSE) metric to get a value of 0.003 and using Quadratic Weighted Kappa (QWK) to get a value of 0.92. The results of this evaluation conclude that the assessment produced by the model is quite accurate and in line with the standard manual assessment. With this research, it is hoped that it can be a good start in developing an automatic essay grading system based on Indonesian and can be implemented for various purposes in the field of education.