ETD Recommendation System

Abstract

Our project involves expanding upon a previous recommendation system built by CS 5604 students. Previous CS 5604 teams have created a chapter summarization model to generate summaries for over 5000 Electronic Theses and Dissertations (ETDs). We used these summaries to fuel our recommendation system. Using chapter summaries improved our ability to predict resources that a user may be interested in because we narrowed our focus to individual chapters rather than the abstract of the whole paper. Authors will benefit from this recommendation system because their work will be more accessible. We provide a web page for users to explore how different clustering algorithms impact the search results, giving the user the ability to modify parameters such as the number of clusters and minimum cluster size. This web page will appeal to niche users interested in experimenting with recommendation systems, allowing them to fine-tune the recommendation results. We recommend for future work to continue exploring different clustering algorithms, as well as using our chapter recommendations to fuel a recommendation list based on each chapter. During this project, we learned about clustering algorithms, working as a team, and starting a project from the ground up. A previous CS5604 team built a stand-alone website that supports search, a recommendation system, and the ability to experiment with different search methods. During this semester, we expanded upon the existing website, using clustering algorithms to experiment with the recommendation system. Users may specify different parameters to understand how different clustering algorithms may change the recommendations.