Implementing Similarity Matrices in a Web Interactive Tool for Scientific Articles Matching

Due to the constant increase in the publication of scientific articles in databases worldwide, searching for relevant academic documents for research has become a complex process that sometimes requires significant time and effort. This research article addresses this problem by developing an interactive web system that makes it easier for users to identify the similarity of scientific articles for their research. The system allows users to upload a dataset of scientific articles with three attributes, titles, keywords, and abstracts, to analyze their similarities. The methodology uses NLP techniques for pre-processing and homogenization of the attributes. Then, we calculated similarity matrices using two metrics that assemble the three attributes into a single weighted matrix to measure the divergence between scientific documents. The results are presented through a web interface, including summary tables, heat maps, two-dimensional scatter diagrams, and clustering graphs. The experiment was carried out using a dataset of 192 scientific articles from different academic databases covering the areas of formal sciences, computer sciences, medicine, and social sciences. The results revealed more significant similarities between the documents in the same study field.