Implementasi Term Frequency dan Cosine Similarity pada Aplikasi Koreksi Otomatis Jawaban Esai Berbasis Web
TL;DR
Term Frequency and Cosine Similarity can be implemented as a mechanism for measuring textual similarity in web-based essay grading applications and validated using a larger dataset of student answers and comparisons with teacher assessments is required to evaluate the method's performance more comprehensively.
Abstract
Manually grading essay answers is time-consuming and labor-intensive, as teachers must review each student's response individually against an answer key. This study aims to implement Term Frequency (TF) and Cosine Similarity within a web-based automated essay grading application. The system was developed using the Waterfall Software Development Life Cycle (SDLC) model, with text processing involving case folding, tokenization, filtering, and stemming stages. Preprocessing results were represented as term frequencies and subsequently compared using Cosine Similarity. A sample calculation using a single pair—an answer key and a student response—yielded a Cosine Similarity score of 0.878 (87.8%), indicating the level of textual similarity between the two documents; this figure does not represent the system's overall accuracy, as algorithm testing in this study was limited to a single document pair and did not include a comparison with teacher assessments. Functional testing using Black Box Testing demonstrated that the application's core functions operated according to the designed scenarios. The results indicate that TF and Cosine Similarity can be implemented as a mechanism for measuring textual similarity in web-based essay grading applications. Validation using a larger dataset of student answers and comparisons with teacher assessments is required to evaluate the method's performance more comprehensively.
