what-when-how
In Depth Tutorials and Information
Last, identify the particular weight change of interest words caused by the used
patterns of bloggers.
Interest words are defined as words related to a blogger's interest. For example,
if a blogger's interest is computer, the interest words are “computer,” “hardware,”
etc. To compare the differences between interest words, use the interest word ratio
as an indicator. Interest word ratio is defined as the ratio of the number of interest
words to the number of words in all weblogs. Similarly, an interest post is defined
by the weblog entry related to a blogger's interest, and interest post ratio indicates
the percentage of the number of interest posts to all posts. Tf-idf is adopted as the
term-weighting scheme. he weight change, that is, the change of term weight, or
the ratio of the new term weight to the average term weight, is an indicator of the
interest words used by interested bloggers compared to the average bloggers.
Observing the relationship between the interest word (post) and time based on
real data, two phenomena were found: (1) interested bloggers post interest posts/
words more frequently than uninterested bloggers, and (2) the frequency of interest
posts/words posted by interested bloggers does not change significantly with time.
hat is, the result implies that interested bloggers post more regularly than unin-
terested bloggers. Two metrics are used to measure these two phenomena: average
time period between two interest posts (formula 3.16), and variance of the time
period between two interest posts.
=
∑
Period between post
Number of pos
Period per post
(3.16)
t
Interactive features represent the degree of interactivity at which a blogger acts in
the blogosphere. Interactive features consist of response time (formula 3.17), length
of the comments, and the frequency of interest-related comments. he higher the
degree of interactivity, the higher the probability that the blogger may have this
interest.
Response time
=
comment time
−
post time
(3.17)
A machine-learning approach can be applied to detect users' interests using these
three features.
Besides detecting users' interests, let us also investigate the influence factors
of users' interests and proposed models to predict the tendency of users' inter-
ests. Santos-Neto et al. [40] analyzed whether usage patterns can be harnessed
to improve navigability in a growing knowledge space. he author studied col-
laborative tagging social networks with CiteULike and Bibsonomy, including
presenting a formal definition of tagging communities, characterizing tagging
activity distribution among users, and investigating the structure of users' shared
interests. From this they could define the interest-sharing graph and investigate
