Projects > Spectrum


Spectrum is a meta-search system for blogs that enables users to find opinions from bloggers interested in different domains (e.g. business, politics, and religion). Using the system, users can search for blogs from different perspectives about the same topic. For example, for the issue of "abortion", the system finds blogs related to religion, law and health, etc. Blogs from religious blog sites discuss "abortion" in the context of various religious beliefs. In contrast, law bloggers discuss legislation related to abortion from a legal point of view. Health care bloggers posts practical information about abortion choices.

To categorize the interests of bloggers, we develop a two-layer classification model. The model classifies the bloggers' interests based on short snippets of their blog posts. Experiment results show that the model is robust to the noise in text classification of blog posts and achieves satisfactory performance.

Project Papers

  • Categorizing Blogger's Interests Based on Short Snippets of Blog Posts