Proposed Model of Academic Reading Material Recommendation System

Cold-start problem and cold-item are something that will happen when an early developed online library of educational institution library tries to recommend scientific articles to users. The reading materials do not even have reviews and/or ratings from previous users, no users have expressed preferences yet, also solely rely on keywords in search engines. The fact that there are abundant holdings in the library, it needs to effectively maintain users' interests to borrow and download academic reading material in accordance with users' interest from holdings in the library repository. This study seeks to provide novelty by finding another way to utilize dataset with only using abstract and title variables as an input parallelly that can provide effective results as a recommendation system. It proposes a word embedding model to be used as topic modeling for the content-based recommendation system to overcome the problems, wherein the attributes are minimum (such as title, author, and abstract) and user data are not available.