ScienceWISE: Topic Modeling over Scientific Literature Networks

TL;DR

This paper presents methods to improve topic modeling and article recommendation in the ScienceWISE system by addressing network partitioning challenges through concept identification, bipartite network representation, and clustering algorithms.

Contribution

It introduces three novel research directions to enhance the automatic assignment of articles to research topics in large scientific literature networks.

Findings

Enhanced inter-article similarity through generic concept identification

Improved scalability using bipartite network representation

Effective clustering for cross-disciplinary and fine-grained topics

Abstract

We provide an up-to-date view on the knowledge management system ScienceWISE (SW) and address issues related to the automatic assignment of articles to research topics. So far, SW has been proven to be an effective platform for managing large volumes of technical articles by means of ontological concept-based browsing. However, as the publication of research articles accelerates, the expressivity and the richness of the SW ontology turns into a double-edged sword: a more fine-grained characterization of articles is possible, but at the cost of introducing more spurious relations among them. In this context, the challenge of continuously recommending relevant articles to users lies in tackling a network partitioning problem, where nodes represent articles and co-occurring concepts create edges between them. In this paper, we discuss the three research directions we have taken for solving this issue: i) the identification of generic concepts to reinforce inter-article similarities; ii) the adoption of a bipartite network representation to improve scalability; iii) the design of a clustering algorithm to identify concepts for cross-disciplinary articles and obtain fine-grained topics for all articles.