Perbandingan Kesamaan Tugas Mahasiswa Berbasis Text Summarization Menggunakan Metode Cosine Similarity

Abstract

The manual process of checking student assignments for similarities can be time-consuming and labor-intensive. Implementing text summarization allows for the extraction of important information from lengthy student assignment texts, enabling the identification of similarities in submitted assignments. Therefore, this study applies text summarization methods to reduce the length of document answers, and then compares the summary results to expedite the assessment process. The data used was obtained from assignment archives of a particular course, consisting of 90 documents with 4 essay questions. Summarization is performed by ranking word weights generated using TF-IDF weighting according to the highest weights. The summary results are then compared using cosine similarity. The research results indicate that the system is capable of generating summaries consisting of the highest-weighted words, with evaluation results showing an accuracy of 94.4%. This means that the compared summaries have a fairly high degree of similarity. Meanwhile, the document similarity evaluation by experts shows that out of 105 data comparisons, 67 were found to be consistent, equating to 63.80%. This discrepancy is due to the system only comparing based on the words present in the summary, not based on their meaning.