Tadanobu Furukawa, Yutaka Matsuo, Ikki Ohmukai, Koki Uchiyam, Mitsuru Ishizuka
We propose a ranking technique which extracts topics and innovators by analyzing that process. Our method identifies URLs of topics and the bloggers who trigger topic diffusion. Our assumption is that the strength of propagation of a topic is determined by the influences of topics and bloggers. This decomposition is attained through singular value decomposition (SVD); We construct a diffusion matrix: the first left and right singular vectors are regarded respectively as the influences of the topics and the bloggers. We show that our method can extract propagative topics (which is not bursty because of the media effect) as well as the influential bloggers.
Subjects: 1.10 Information Retrieval; 1. Applications
Submitted: Feb 15, 2008