Relevancy Prediction of Scientific Articles Using Similarity Measures and Citation Mention Rate

To guarantee that scientific areas continue to evolve effectively, it is important to acknowledge the scientific contribution of earlier research work. Citations play a vital part in the ranking of authors, journals, institutions, and organizations. Scientific works, for many reasons, need to contain suitable citations for other works. In recent works, researchers have only focused on identifying important citation by using classifier to classify important and non-important classes. They used feature selection methods to select unimportant features for classifier. This work intends to propose a system for predicting the relevancy of citing and cited articles based on both similarity measures and citation mention rate by using relevancy prediction method without classifier. Firstly, the contents and references metadata are extracted by using the Cermine tool. In second, the inline citation extraction and matching are performed using citation anchor taxonomy. Finally, the relevancy between citing and cited papers are predicted based on references, content, and citation-content similarities. These similarities are calculated using Bag-of-Words model and Cosine Similarity. After that, the relevance of a cited work to a citing paper based on similarity scores and citation-mention rate are examined. The proposed work uses four scientific articles that cited a target set of 20 papers. The results reveal that the proposed approach achieved higher similarity scores based on the relationship between citing and cited papers.