AUTHOR-CENTRIC SCIENTIFIC PAPER RECOMMENDER SYSTEM TO IMPROVE CONTENT-BASED FILTERING APPROACH

Scholarly publications on the web are rapidly expanding, making it difficult for scholars to identify relevant study materials. Information overload makes it harder to find important material, especially for new researchers. Scholarly recommender systems solve this issue by employing recommendation techniques to assist researchers in locating appropriate literature based on their interests. Existing systems frequently rely on user profiles and public and non-public metadata, which leads to the persistent problem in scholarly recommendations called cold start. To deal with the challenges of cold start in scholarly-based recommender systems, this research suggests an improved Content-Based Filtering (CBF) approach that takes advantage of publicly available metadata, specifically the author(s) feature. The approach incorporated author(s) features into a scholarly recommender system to serve as a basis and key component for alleviating "A New Paper Cold Start Problem." The proposed approach implements the feature vectors of the metadata using the Count Vectorizer and similarity computation was performed using the Cosine Similarity formula."  An experiment using a publicly available dataset shows that the suggested method surpasses the approaches previously proposed by other researchers regarding recommendation accuracy and relevancy, making it a dependable and efficient instrument for scholarly recommendation. The result also shows the effectiveness of the author(s) feature in tackling new papers in scholarly recommendation systems.