ABSTRAK
Algoritma Winnowing merupakan algoritma dengan sifat character-based, dimana algoritma tersebut menggunakan fingerprint dalam mendeteksi tingkat kesamaan antar teks. Skripsi ini membahas mengenai penerapan dan pengembangan sistem penilai esai otomatis berbasis algoritma winnowing, serta metode-metode untuk mengukur tingkat kesamaan fingerprint. Pengujian beberapa pendekatan metode dilakukan untuk meningkatkan tingkat akurasi dari sistem. Pendekatan metode tersebut adalah dengan metode cosine similarity, jaccard index, dan dice similarity. Implementasi dari algoritma winnowing berhasil meningkatkan akurasi sistem. Peningkatan akurasi ini diukur dengan membandingkan SIMPLE-O berbasis winnowing dengan SIMPLE-O yang telah dikembangkan sebelumnya yaitu berbasis LSA Latent Semantic Analysis .

---

ABSTRACT
Winnowing algorithm is a character based algorithm. Winnowing used fingerprint to detect the similarity between texts. This thesis discusses the application and the development of automated essay grading SIMPLE O based on winnowing algorithm, as well as methods to measure the similarity between fingerprints. Several text similarity approaches have been tested to improve the accuracy of the system. In term of text similarity, cosine similarity, jaccard index, and dice similarity are used to measure the similarity between texts. Stemming and Synonym Recognition also implemented in this system. Winnowing algorithm has improved the accuracy of the system. The accuracy was measured by comparing SIMPLE O winnowing based with SIMPLE O LSA Latent Semantic Analysis based.