Citation Analysis on Scientific Articles Using Cosine Similarity

Citations have a vital role in scientific articles. Every sentence that is listed must be obtained from the reference. There is a measure that is similarity. The problem is, the likeness of a sentence in a scientific article to the source it refers to can have different meanings. The high level of similarity can undoubtedly increase the similarity value if it is checked using the plagiarism (similarity checker) tool, where the higher the value is more similar to other articles on the contrary with low similarity remaining. However, a low level of similarity can also mean that a statement in a scientific paper has referred to a completely irrelevant source. This research aims to analyze the level of similarity between scientific articles and other articles or sources it refers to. Checking the input text starts with text preprocessing, weighting the TF-IDF, and determining the similarity level using Cosine Similarity. Based on the test results, it is found that the results in a scientific article document have a high level of similarity with the article it refers to, and the other 30% have a low level of similarity, or in other words, there is no relationship.