Empirical Analysis of Periodically Bookmarked Web Pages in Social Bookmarking Systems

There are enormous amount of pages on web. We consider that there are many web pages periodically viewed by users. For example, we can imagine that pages about the illustration of New Year's cards are generally viewed near the end of each year. In this paper, we investigate the property of such periodically viewed web pages using social book marking systems. Specifically, we will confirm that how many pages are actually book marked with a specific period and why they are book marked periodically. We focus on the frequency analysis with a discrete Fourier transform as a concrete tool to analyze the temporal periodicity of bookmarks, and we use the Hatena Bookmark dataset as a concrete dataset of social bookmarks. As a result, we detected 1385 periodically book marked pages which accounts for approximately 1% of the pages stored in the dataset.