Matching Scientific Article Titles using Cosine Similarity and Jaccard Similarity Algorithm

Abstract

This study compared various methods for academic article similarity matching. We employed two similarity algorithms, specifically Cosine and Jaccard Similarity. Moreover, these two similarity algorithms were combined with TF-IDF to increase the similarity results. We performed 16 experimental scenarios, measured based on the processing speed and accuracy, to compare the two algorithms. Highest similarity value in Cosine Similarity with circumstances without vectors, either with or without stemming. The findings showed that Cosine similarity performs better than Jaccard similarity.