| Web Mining Project Homepage |
Welcome to our project web page for CSCI B659, Web Mining, for Spring 2005.
Traditional search engines provide links to pages - perhaps in conjunction with a few extracted sentences - in response to a query. We propose a system that presents search results over a blog corpus as a summarization of the information in posts. We compute this summarization using the MEAD multidocument summarization system. We evaluate our results by comparing them to a summarization computed by a simpler method, and to unsummarized results; the method for evaluation is a user study.
Alex Breuer, abreuer (at) cs (dot) indiana (dot) edu
Jacob Ratkiewicz, jpr (at) cs (dot) indiana (dot) edu
Check out all our latest news in the exciting blog, or read our amazing paper.
One of our most interesting discoveries was the consistently strong performance of our naïve summary. The following graphs show the performance of this summarizer in the user study; note that the naïve summarizer received votes in almost every query. This suggests to us that while the naïve summarizer isn't always best, it is always good.
