Pradityo, Muhammad Hilmy Adhi (2025) Automatic Essay Scoring Untuk Menilai Kemampuan Berpikir Kritis Pada Esai Bahasa Indonesia: Studi Kasus Soal Fisika SMA Tentang Dampak Gas Rumah Kaca. Other thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 5027211053-Undergraduate_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
5027211053-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy |
Abstract
Penilaian esai secara manual memiliki tantangan seperti subjektivitas penilai dan proses evaluasi yang memakan waktu. Dalam pembelajaran Fisika SMA, esai digunakan untuk menilai kemampuan berpikir kritis siswa, namun penilaian konvensional sering tidak konsisten dan kurang efisien. Oleh karena itu, penelitian ini mengembangkan sistem Automatic Essay Scoring (AES) berbasis Natural Language Processing (NLP) menggunakan kombinasi IndoBERT dan BiLSTM. IndoBERT digunakan sebagai representasi teks, sementara BiLSTM digunakan untuk klasifikasi skor. Sebagai metode klasifikasi pembanding, SVM dan IndoBERT juga digunakan dalam penelitian. Hasil penelitian menunjukkan bahwa model BiLSTM dan SVM memiliki performa yang hampir serupa dalam enam kategori penilaian berpikir kritis. Model BiLSTM mencapai akurasi tertinggi pada kategori reason (0,69), situation (0,85), dan clarity (0,65), sedangkan model SVM menunjukkan akurasi tertinggi pada kategori focus (0,64) dan overview (0,80). Sementara itu, model IndoBERT menghasilkan performa terbaik pada kategori inference (0,70). Uji coba implementasi sistem menunjukkan bahwa website dapat memproses file dan menampilkan hasil penilaian secara otomatis dalam waktu rata-rata 5 detik. Website yang dibangun juga mendapat respons positif dari pengguna, dengan skor kepuasan dan kejelasan hasil sebesar 5 dari 5. Sistem AES ini dinilai mampu melakukan penilaian esai dengan cepat, objektif, dan konsisten.
==========================================================================================================================================
Manual assessment of essays has challenges such as subjectivity of the assessor and a time-consuming evaluation process. In high school Physics learning, essays are used to assess students' critical thinking skills, but conventional assessments are often inconsistent and less efficient. Therefore, this research develops a Natural Language Processing (NLP)-based Automatic Essay Scoring (AES) system using a combination of IndoBERT and BiLSTM.
IndoBERT is used as text representation, while BiLSTM is used for score classification. As a comparison classification method, SVM and IndoBERT were also used in the study. The results showed that the BiLSTM and SVM models performed almost similarly in the six categories of
critical thinking assessment. The BiLSTM model achieved the highest accuracy in the reason (0.69), situation (0.85), and clarity (0.65) categories, while the SVM model showed the highest accuracy in the focus (0.64) and overview (0.80) categories. Meanwhile, the IndoBERT model produces the best performance in the inference category (0.70). The system implementation trial shows that the website can process files and display assessment results automatically in an average of 5 seconds. The website also received a positive response from users, with a satisfaction score and clarity of results of 5 out of 5. This AES system is considered capable of conducting essay assessments quickly, objectively, and consistently.
Actions (login required)
 |
View Item |