TREC 2006 Blog Track Notebook
Blog Track Open Task: Spam Blog Classification
November 14, 2006
Spam blogs or Splogs are blogs created for the sole purpose of hosting ads, promoting affiliate sites and getting new content indexed, with auto-generated or plagiarized content from other sources. Spammers equipped with readily available splog creation software inundate the blogosphere both at ping servers, and at systems that index and analyze blogs. Our own studies estimate these numbers to be around 75% at ping servers and 20% at popular blog search engines. In this open submission we hence propose Spam Blog Classification as a new task in the Blog Track. Splogs are a specific instance of the more general spam web-pages. While offline graph based mechanisms like TrustRank are quite effective and sufficient for the Web, the blogosphere demands new techniques. The quality of blog search engines is judged not just by their reach, but also by their ability to index recent (non-spam) posts. This requires that fast online splog detection/filtering be used prior to indexing new content, followed by offline techniques that exploit link graph anomalies. The nature of this problem makes splog detection challenging. This open task submission underscores the seriousness of the splog problem in the TREC 2006 collection, details how it impacts the primary task of Opinion Identification, and proposes multiple assessment and evaluation approaches for a Spam Blog Classification task in Blog Track 2007.
InCollection
Downloads: 8579 downloads
Google Scholar Citations: 11 citations