Implementasi Algoritma Cosine Similarity Berbasis Natural Language Processing Pada Penilaian Esai Otomatis

Naya Sabilla Putri; Alan Novi Tompunu; Adi Sutrisman

doi:10.33365/jtk.v20i2.2422

Naya Sabilla Putri Politeknik Negeri Sriwijaya
Alan Novi Tompunu Politeknik Negeri Sriwijaya
Adi Sutrisman Politeknik Negeri Sriwijaya

DOI: https://doi.org/10.33365/jtk.v20i2.2422

Keywords: Cosine Similarity, Natural Language Processing (NLP), Penilaian Esai Otomatis, Computer Based Test (CBT), Mean Absolute Error (MAE).

Abstract

Abstrak− Evaluasi tes esai berbasis kertas di SMP Srijaya Negara menghadapi kendala durasi koreksi yang lama, risiko salah hitung, serta bias subjektivitas. Demi mengatasi persoalan tersebut, dikembangkan sistem Computer Based Test (CBT) berbasis web menggunakan teknologi Natural Language Processing (NLP) dan algoritma Cosine Similarity untuk menilai esai otomatis secara objektif. Pengembangan aplikasi ini menerapkan metodologi Rapid Application Development (RAD) dengan kombinasi pemrograman PHP, framework Laravel, dan database MySQL. Dalam prosesnya, modul NLP mengeksekusi empat tahapan preprocessing: case folding, tokenizing, filtering, dan stemming. Dokumen hasil pembersihan kemudian dikonversi menjadi vektor numerik melalui skema Bag of Words (BoW) untuk dihitung skor kemiripannya berdasarkan sudut kosinus antarvektor. Validasi keakuratan sistem diuji menggunakan 30 sampel data jawaban riil siswa Kelas VIII pada mata pelajaran Ilmu Pengetahuan Alam (IPA). Tingkat presisi model diukur menggunakan metrik Mean Absolute Error (MAE) dengan menyandingkan skor otomatis sistem terhadap nilai aktual dari guru. Hasil pengujian menunjukkan sistem berhasil memperoleh nilai MAE sebesar 3,87 dari skala 0–100. Angka kesalahan di bawah deviasi 5,00 ini membuktikan bahwa algoritma Cosine Similarity memiliki tingkat akurasi tinggi dan layak diandalkan sebagai instrumen bantu evaluasi akademik di sekolah.

Kata Kunci: Cosine Similarity, Natural Language Processing (NLP), Penilaian Esai Otomatis, Computer Based Test (CBT), Mean Absolute Error (MAE).

Abstract− The implementation of conventional paper-based essay examinations at SMP Srijaya Negara encounters operational constraints, particularly regarding prolonged grading durations, calculation error risks, and subjective biases. To address these issues, a web-based Computer-Based Test (CBT) platform featuring automated essay grading was designed and engineered utilizing Natural Language Processing (NLP) and the Cosine Similarity algorithm to ensure objective text matching. Developed under the Rapid Application Development (RAD) framework to optimize efficiency, this system runs on PHP, the Laravel framework, and a MySQL database backend. For text manipulation, the NLP module executes four primary preprocessing stages: case folding, tokenizing, stopword filtering, and stemming. The normalized texts are subsequently mapped into numerical vectors using the Bag of Words (BoW) method to determine similarity scores based on the cosine angle between vectors. System accuracy was validated utilizing 30 authentic eighth-grade science essay responses, evaluated via Mean Absolute Error (MAE) statistics against manual teacher grades. The empirical results yielded an MAE value of 3.87 on a 0–100 scale. This minimal error rate, remaining well below the 5.00 deviation threshold, confirms that the Cosine Similarity algorithm delivers exceptional precision and marginal variance when assessing essay-based responses. Consequently, the constructed CBT system serves as a highly feasible, valid, and dependable auxiliary instrument to support academic evaluation efficiency in schools.

Keywords: Cosine Similarity, Natural Language Processing (NLP), Automated Essay Grading, Computer Based Test (CBT), Mean Absolute Error (MAE).

References

M. Maulani, “RANCANG BANGUN APLIKASI COMPUTER BASED TEST BERBASIS WEB PADA SMPN 1 KATAPANG KABUPATEN BANDUNG SELATAN,” 2021.

D. Setiawan, “Penerapan Exambro Sebagai Pendukung CBT (Computer Base Test),” Nov. 2021.

D. Wijayanto, “Manajemen Evaluasi Pembelajaran Computer Based Testing (CBT) di SMP Kapanewon Wonosari Gunungkidul,” Oct. 2022, [Online]. Available: http://jurnal.ustjogja.ac.id/index.php/mmp

I. P. Y. Agus Ariwanta, I. G. T. Eka Saputra, N. P. E. Apriyanthi, I. M. A. O. Gunawan, and G. Indrawan, “Analisis Kepuasan Pengguna Menggunakan Metode EUCS Pada Sistem Computer Based Test di Institusi Pendidikan,” Journal of Computer System and Informatics (JoSYC), vol. 4, no. 4, pp. 942–951, Aug. 2023, doi: 10.47065/josyc.v4i4.3752.

F. Rumaisa, Y. Puspitarani, A. Rosita, A. Zakiah, and S. Violina, “Penerapan Natural Language Processing (NLP) Di Bidang Pendidikan,” Dec. 2021.

R. R. Rafi, R. A. Ramadhani, and A. Sanjaya, “PEMANFAATAN ALGORITMA COSINE SIMILARITY UNTUK MENGKOREKSI UJIAN ESAI,” JIPI (Jurnal Ilmiah Penelitian dan Pembelajaran Informatika), vol. 10, no. 3, pp. 2242–2254, Aug. 2025, doi: 10.29100/jipi.v10i3.8058.

F. Ri. Kurniawan and F. M. Arif, “IMPLEMENTASI TEXT MINING MENGGUNAKAN METODE COSINE SIMILARITY UNTUK KLASIFIKASI KONTEN BERITA DI POSTINGAN GRUP FACEBOOK INFO LANTAS DAN KRIMINAL PASURUAN,” JAMI: Jurnal Ahli Muda Indonesia, vol. 3, no. 1, pp. 9–17, Jun. 2022, doi: 10.46510/jami.v3i1.41.

N. Rokhman, P. A. Maulan, and N. A. Wirahuda, “Analisis Penilaian Esai Secara Otomatis Menggunakan Natural Language Processing (NLP) dan Cosine Similarity,” Go Infotech: Jurnal Ilmiah STMIK AUB, vol. 31, no. 1, pp. 41–52, Jun. 2025, doi: 10.36309/goi.v31i1.359.

Y. Natadisastro, N. Safaat Harahap, and T. Darmizal, “MALCOM: Indonesian Journal of Machine Learning and Computer Science Implementation of Large Language Models in the Design of Android-Based Computer-Assisted Test Mobile Applications Implementasi Large Language Models dalam Rancang Bangun Aplikasi Computer-Assisted Test Mobile Berbasis Android,” vol. 6, no. 1, pp. 52–61, Jan. 2026, doi: 10.57152/malcom.v6i1.2402.

A. Basri, V. Kuswanto, and A. Leo, “Rancang Bangun Bridging Sistem Pendaftaran dan Aplikasi Mobile Jaminan Kesehatan Nasional (JKN),” 2022.

H. Rianto and O. Pahlevi, “Pengembangan Billing Online Sistem dengan Menggunakan Metode Rapid Application Development,” JURIKOM (Jurnal Riset Komputer), vol. 9, no. 6, p. 2021, Dec. 2022, doi: 10.30865/jurikom.v9i6.5213.

D. O. Sihombing, “Implementasi Natural Language Processing (NLP) dan Algoritma Cosine Similarity dalam Penilaian Ujian Esai Otomatis,” Jurnal Sistem Komputer dan Informatika (JSON), vol. 4, no. 2, p. 396, Dec. 2022, doi: 10.30865/json.v4i2.5374.

A. A. Firdaus, A. Id Hadiana, and A. K. Ningsih, “Klasifikasi Sentimen pada Aplikasi Shopee Menggunakan Fitur Bag of Word dan Algoritma Random Forest,” R2J, vol. 6, no. 5, 2024, doi: 10.38035/rrj.v6i5.

R. P. Permata, R. N. Suharto, L. Candra, and A. Julianty, “Towards an Automated Essay Evaluation System NLP Based Text Embeddings and Similarity Metrics,” Jurnal Teknologi Informasi & Komunikasi, vol. 16, Apr. 2025, doi: 10.31849/digitalzone.v16i1.

A. Arsad, M. Hamid, and M. Santosa, “PENERAPAN TEKS MINING DAN COSINE SIMILARITY UNTUK MENENTUKAN KESAMAAN DOKUMEN SKRIPSI APPLICATION OF TEXT MINING AND COSINE SIMILARITY TO DETERMINE THE SIMILARITY OF THESIS DOCUMENTS,” Apr. 2024.

A. M. Fawwazrin, A. Arbansyah, and Amin Rezaeipanah, “ID Card Printing Time Prediction With Mean Absolute Error in Linear Regression,” JSE Journal of Science and Engineering, vol. 3, no. 2, pp. 80–84, Jan. 2025, doi: 10.30650/jse.v3i2.3868.

C. Binangkit, A. Voutama, and N. Heryana, “PEMANFAATAN UML (UNIFIED MODELING LANGUAGE) DALAM PERENCANAAN SISTEM PENGELOLAAN SEWA ALAT MUSIK BERBASIS WEBSITE,” 2023.

R. P. Pangestu and A. Voutama, “PEMANFAATAN UML (UNIFIED MODELLING LANGUAGE) PADA SISTEM PENGELOLAAN ASPIRASI MAHASISWA BERBASIS WEBSITE,” 2024.

P. Putra, V. Pranatawijaya, N. Sari, and Widiarty, “Mobile Android Queue Covid19 Application Registration and Application of Queues at the Mobile,” vol. 16, no. 1, 2022, doi: 10.47111/JTI.

F. Ardhy et al., “Pelatihan Analisis dan Desain Sistem Informasi Menggunakan Unified Modeling Language (UML) di SMK Pelita Madani Kabupaten Pringsewu,” Abdimas Universal, vol. 5, no. 1, pp. 97–104, Apr. 2023, doi: 10.36277/abdimasuniversal.v5i1.285.