Research Paper Recommendation System

Subodh Gholve


Supervised by Luis Espinosa-Anke; Moderated by Alun D Preece

Covid-Aware, the paper recommendation tool is developed to help the researchers in finding the research papers most relevant to their paper of interest. This does not use user’s evaluation and the recommendations are calculated using the information within the dataset of the chosen paper. The recommender is developed using principles of content-based filtering, one of the most common approaches to recommender systems. To achieve this, I have used Cosine similarity function from Scikit-learn machine learning libraries in Python, Flask as backend and a simple HTML frontend to display the results. The dataset being used is a cleaned version of the CORD-19 Open Research Dataset provided by Semantic Scholar for use by global researchers. This dataset contains Abstract IDs and Paper Abstracts for the research papers. The recommendation tool takes an Abstract ID as input to compare the respective Paper Abstract with the rest of the dataset to find the Top 10 most relevant documents.

Final Report (29/12/2020) [Zip Archive]

Publication Form