Abstract:Academic search engines have been widely used to access academic papers, where users’ information needs are explicitly represented as search queries. Some modern recommender systems have taken one step further by predicting users’ information needs without the presence of an explicit query. In this article, we examine an academic paper recommender that sends out paper recommendations in email newsletters, based on the users’ browsing history on the academic search engine. Specifically, we look at users who regularly browse papers on the search engine, and we sign up for the recommendation newsletters for the first time. We address the task of reranking the recommendation candidates that are generated by a production system for such users. We face the challenge that the users on whom we focus have not interacted with the recommender system before, which is a common scenario that every recommender system encounters when new users sign up. We propose an approach to reranking candidate recommendations that utilizes both paper content and user behavior. The approach is designed to suit the characteristics unique to our academic recommendation setting. For instance, content similarity measures can be used to find the closest match between candidate recommendations and the papers previously browsed by the user. To this end, we use a knowledge graph derived from paper metadata to compare entity similarities (papers, authors, and journals) in the embedding space. Since the users on whom we focus have no prior interactions with the recommender system, we propose a model to learn a mapping from users’ browsed articles to user clicks on the recommendations. We combine both content and behavior into a hybrid reranking model that outperforms the production baseline significantly, providing a relative 13% increase in Mean Average Precision and 28% in Precision@1. Moreover, we provide a detailed analysis of the model components, highlighting where the performance boost comes from. The obtained insights reveal useful components for the reranking process and can be generalized to other academic recommendation settings as well, such as the utility of graph embedding similarity. Also, recent papers browsed by users provide stronger evidence for recommendation than historical ones.

Personalized Paper Recommendation in Online Social Scholar System.

A personalized paper recommendation method considering diverse user preferences

Personalized Paper Recommendation Based on User Historical Behavior

Personalized Academic Research Paper Recommendation System

Personalised Reranking of Paper Recommendations Using Paper Content and User Behavior

Personalized recommendation based on implicit social network of researchers

SSRES: A Student Academic Paper Social Recommendation Model Based on a Heterogeneous Graph Approach

A Personalized Researcher Recommendation Approach in Academic Contexts: Combining Social Networks and Semantic Concepts Analysis.

Paper Recommendation Based on Author-paper Interest and Graph Structure

Method for Scientific Paper Recommendation Based on Academic Community

Deep Learning-based Personalized Paper Recommender System

RECENT ADVANCES IN PERSONAL RECOMMENDER SYSTEMS

Recommending academic papers via users' reading purposes.

Personalized Recommender Systems Based on Social Relationships and Historical Behaviors

Analysis and Design of a Personalized Recommendation System Based on a Dynamic User Interest Model

A hybrid recommendation system for researchgate academic social network

Personalized Recommendation Leveraging Social Relationship

Research Of Personalized Recommendation System Based On Learning Demand Of Mining

A Recommendation System for Paper Submission Based on Vertical Search Engine

Survey of Personalized Learning Recommendation

Supporting Undergraduate Research: Recommending Personalized Research Projects to Undergraduates