Page 112 - 臺大管理論叢第32卷第1期

Internet Celebrity Economy: Exploring the Value of Viewers’ Comment Features and Live Streamers’
Marketing Strategies in Forecasting Revenue

the first comment and gift posted to the last ones posted after the live stream finished.
We chose Douyu over other possible platforms because of the following reasons: (1)
paid gifting is originally invented by Chinese companies (Zhou et al., 2019) and has been
widely adopted by Chinese live-streaming firms. (2) DouYu is the most top live-streaming

platforms in China as mentioned previously (Tan, 2019). DouYu had approximately 46.71
million active users in January 2019. The platform focuses on streaming such content as
5
food, mobile games, travel, music, cars, science, sports, and other recreational activities. (3)

DouYu requires no access fee for either browsing the live streams or posting a comment;
therefore, it can attract more viewers, and the research data will be more diverse. (4) The
DouYu website is well designed so that information collection is straightforward, thus
reducing error during data collection. Thus, it is possible to track the timepoint in the live
stream's run to which a particular comment belonged.

As mentioned previously, Zhou et al. (2019) measure the numbers of donations re-
ceived by streamers and the comment features of viewers at the minute level during a
live-streaming period. However, since viewer donations are partly dependent on what

streamers air, it is logical to consider the streamers’ behaviors when examining the reasons
for donations. In terms of collecting data on streamers’ behaviors, this study downloads
each live-streaming video and performs content analysis for each streamer behavior. In ad-
dition, Zhou et al. (2019) only use the fixed time span of “one minute” as the analysis unit,
which cannot effectively identify the distribution patterns hidden in streamer behaviors

and gift-sending information. One reason is that it is not reasonable to divide the stream-
ers’ behaviors into units per minute. Further, the viewers’ behaviors are mainly affected by
streamers’ behaviors, and viewers’ comments or donations are the results of their lengthy

considerations; a viewer’s behavioral state will have a carryover effect. Therefore, to cap-
ture fully the relationship between the behaviors of the streamer and those of the viewers,
this study uses seconds as the time span unit and observes the correlation between the
streamer behaviors and viewer behaviors as a basis for correction. Specifically, this study
codes 1 at five seconds before the viewer behavior occurs (such as sending a comment or

donation), ten seconds after the viewer behavior occurs as well as ten seconds after the

5 Miss Game (2019)

104