Aggregating, Classifying, and Ranking Research Blogs

Abstract

Keeping up with research and staying at the forefront of one's field is both a crucial and difficult task for researchers. This thesis describes a tool designed to tackle this problem by aggregating and classifying research blogs, which are often more accessible to both readers and writers than formal papers. We build a corpus of over 10,000 research blogs by crawling through blogroll links. To identify communities of related blogs, we explore methods such as clustering algorithms, topic modeling, and network analysis. Ultimately, we group blogs into communities using modularity maximization on a network augmented by semantic similarity scores. Finally, we propose a ranking system for identifying authorities within communities. Comparing the communities identified by this tool with an existing blog aggregator reveals great similarities and demonstrates the potential of this tool for use on a large collection of research blogs.