| We present BlogScope (www.blogscope.net), a system that analyzes the Blogosphere1. BlogScope is an information discovery and text analysis system. It's features include spatio-temporal analysis of blogs, automatic identification of interesting time events as information bursts, generation of topic summaries in form of keyword correlations and burst synopsis, and enhanced ranking functions for improved query answer relevance.; In this thesis, we describe the system by highlighting its information analysis capabilities as opposed to traditional keyword searches. Efficient algorithms for burst identification, correlation discovery, and mining interesting keywords are developed to support various features of the system. We address the problem of aggregating ranked lists as a subproblem in our analysis process. Probabilistic and deterministic algorithms are developed for list aggregation in presence of hierarchies. We describe these algorithms in detail and provide experimental evaluation.; 1Blogosphere is the collective term encompassing all weblogs as a community. |
