Scientific paper recommender system using deep learning and link prediction in citation network.

Area of Science:

• Information Science
• Computer Science
• Bibliometrics

Background:

• The exponential growth of scientific publications necessitates advanced methods for article discovery.
• Traditional content-based and citation-based approaches have limitations in recommending relevant scientific articles.
• Efficiently navigating and discovering pertinent research is a significant challenge for scientists.

Purpose of the Study:

• To design and evaluate a novel recommender system for scientific articles (RECSA).
• To enhance the accuracy and efficiency of scientific article recommendations by integrating content analysis and citation network data.
• To address the limitations of existing methods in handling the increasing volume of scientific literature.

Main Methods:

• Utilizing natural language processing and deep learning (Convolutional Neural Network - CNN) for content analysis of article titles.
• Employing Term Frequency Inverse Document Frequency (TF-IDF) to determine word importance and CNN for feature dimension reduction.
• Calculating content similarity using cosine similarity and analyzing citation networks via link prediction.
• Combining content and citation-based similarity matrices using an influence coefficient for final recommendations.

Main Results:

• The integration of TF-IDF and CNN for content analysis improved precision by at least 0.32% compared to previous methods.
• The RECSA system achieved a first suggestion precision of 99.01%, demonstrating a minimum improvement of 0.9% over compared methods.
• The combined approach of content and citation data significantly enhances recommendation accuracy and efficiency.

Conclusions:

• RECSA offers a more accurate and efficient solution for recommending scientific articles.
• The hybrid approach effectively leverages both textual content and citation network structures.
• This system provides a valuable tool for researchers to navigate the expanding landscape of scientific literature.