Graph-based clustering of documents and words

  • Ajitesh Srivastava
  • Axel Soto
  • Evangelos Milios


In this project we investigate graph-based methods for clustering of documents and words. In a document corpus, the document-term relationship can be modeled as a bipartite graph. One-mode projection of such bipartite graphs can lead to document or word similarity graphs. We propose novel graph-based text clustering algorithms for documents and words. In addition, we propose new metrics for measuring the quality of topics in the form of sets of words extracted from document corpora.