Identifying unusual blog entries: a behavioral approach analyzing blogger's "friends" (グループウェアとネットワークサービス)

概要

論文の詳細を見る
Unusual events in life are often the events we treasure, find interesting and consequently share with friends. We present a behavioral method to identify unusual blog entries within a blog by evaluating comment senders' behaviors. Reader Feedback Vector (RFV) represents each blog entry with a comment sender properties vector and examines each vector for any indication of abnormal comment sender behaviors. An Intimacy Score, which measures the "friendship" level between each comment sender and blogger, is used to weigh each comment sender's properties. This ensures that comment senders who are deemed more "intimate" with the blogger are evaluated more importantly. RFV is subjected to a human-based analysis which asks each test subject to compare the results from RFV to the results from a text-based approach. A second similar test evaluates the significance of Intimacy Score by comparing RFV with and without Intimacy Score (RFV-IS vs. RFV). RFV-IS performed better, by returning more unusual blog entries than its counterpart.
2008-05-16