Users' Behavioral Analysis on Weblogs

This study investigates Weblog users’ behavioral data on a hosting service in Japan. The study analyzes users’ behaviors, such as browsing and bookmarking, in addition to posting comments and sending trackbacks. Results of this study reveal the causes of a user’s visiting and regularly browsing behaviors, especially from a social network point of view. The correlation between the strength of two users on the social network and their visiting behavior is analyzed. Introduction Web logs (blogs) are receiving much attention as a new medium to describe individual experiences and opinions. Blog users sometimes visit others’ blogs and write comments or send trackbacks as they update their own blogs. These users’ behaviors result in relationships among blogs, which create interesting data for analyses of blog communities and topic diffusion. Recently, users have come to read blogs using RSS readers and social networking services (SNSs). We can assume that a user might use social networks to seek blogs. Our analyses use the database of a blog-hosting service in Japan called Doblog1, which NTT Data Corp. and hottolink, Inc. provided. Using this service, users can create and update their blogs easily. They write their blogs, change the templates, and write comments after they log in. Therefore, we can monitor users’ behavior in great detail, even information showing which blogs a user has browsed and when. Doblog has a special function called “bookmark” to link to favorite blogs from one’s own blog. We regard the bookCopyright c © 2006, American Association for Artificial Intelligence (www.aaai.org). All rights reserved. 1 c ©NTT Data Corp., c ©hottolink,Inc., http://www.doblog.com/ , using data of Oct. 2003 – Jul. 2004 mark information as representative of browser bookmarks or blogs of a user’s RSS reader. This study analyzes users’ behaviors from a social network point of view. Three types of relationship are considered: comment, trackback, and bookmark. We call these relationships blog-based relationships. Two types of behavioral relationships considered in this study are: Visit and Regular Reading. We clarify how the blog-based relationship affects users’ behaviors. Mainly, two questions are investigated: Do users’ blog-based relationships affect their browsing behavior? Furthermore, can we predict whether users visit or read a certain blog frequently based on blogbased relation information? Relations between two Blogs Three types of blog-based relations, Comment, Trackback, and Bookmark, are defined as follows: If user UA, who writes blog A, comments on or trackbacks to blog B, we define a relation of Comment / Trackback from A to B. If user UA registers blog B as a bookmark in the Doblog system, we define a relation of Bookmark from A to B. We also define a user’s browsing behavior. If user UA browses B, we call it Visiting behavior from A to B2; if user A has visited user B’s blog regularly, which is defined in this paper as more than 30 times, we call it Regular Reading behavior from A to B. Our hypothesis is that users’ (indirect) relationships affect users’ browsing behavior. Therefore, (i) What relations among Comment, Trackback, and Bookmark behaviors engender Visiting behavior and Regular Reading behavior? We analyzed data of 1647 blogs with top frequent users, who are 10% of all Doblog users. The data consist of 59% of To exclude an unintended browse, we define Visit as the relation by which a user browses more than once. Comment and Trackback relations are defined similarly. all Bookmark data, 64% Comment data, and 64% Trackback data. Relations among two blogs Two blogs sometimes have numerous relations. Because two blogs’ direct relations are explicit and readily apparent, we specifically examine indirect relations between two blogs. A measure of each relational strength is represented by the number of routes that connect two blogs in two hops with the same direction. For example, three routes exist between A and B, as shown in Fig. 1. We can also consider two blogs that are connected by three or more hops, but that relation should be less effective. Moreover, in sociological studies, a two-hop relation is sometimes assessed, e.g., transitivity and clustering coefficients. For that reason, we examine only the number of routes in two hops. Figure 2 shows the extent to which the rate follows the transitivity rule in these two hops. This figure reveals a positive correlation between the number of routes and the transitivity rule. Blog A Blog B blog 1