1.2. PRACTICAL TASKS 3
180,000 people attended with 93% were the young. ird, it delivers an opportunity for remote
counties to advertise their rustic scenery. Daocheng county,
11
a rural town in south-west China,
gained over 10 million likes on TikTok in 2018. In summary, micro-videos are rocking and
taking over the content and social media marketing spaces [120].
Micro-video arises as a new form of user-generated content (UCG) and hence the related
research is relatively sparse. e first work [136] on micro-videos is about the creativity assess-
ment. Redi et al. studied the audio-visual features of creative and non-creative videos. Mean-
while, they introduced a computational framework to automatically predict these categories. In
the same year, Sano et al. [142] observed that the loop is one of the key features of popular
micro-videos, but there are so many non-loop videos mistakenly tagged with “loop.” Inspired
by this observation, they proposed a degree-of-loop method to measure micro-videos. In this
book, we comprehensively work toward analyzing the unique characteristics of micro-videos,
designing theoretical solutions for micro-video understanding, and verifying them over several
practical tasks.
1.2 PRACTICAL TASKS
In this book, we introduce three practical tasks of micro-video understanding, namely popularity
prediction, venue category estimation, and micro-video routing. ese three tasks are leveraged
to verify several state-of-the-art multimodal learning theories we proposed in this book.
1.2.1 MICRO-VIDEO POPULARITY PREDICTION
According to our observation, among the tremendous volume of micro-videos, some popular
ones will be widely viewed and spread by users, while many only gain little attention. is phe-
nomena is similar to many existing social media sites, such as Twitter.
12
For example, one micro-
video about the explosion that interrupted during the France-Germany soccer match in 2015 has
been successfully looped by over 330 million times. Obviously, if we can identify the hot and
popular micro-videos in advance, it will benefit many applications, such as online marketing
and network reservation. Regarding online marketing, the accurate early prediction of popu-
lar micro-videos can facilitate companies planning advertising campaigns and thus maximizing
their revenues. For network service providers, they can timely reserve adequate distributed stor-
age and bandwidth for popular ones, based on the prediction. erefore, it is highly desirable to
develop an effective scheme to accurately predict the popularity of micro-videos.
1.2.2 MICRO-VIDEO VENUE CATEGORIZATION
Organizing micro-videos plays an increasingly pivotal role in high-order analysis of micro-
videos, such as search, browse, and navigation. However, micro-videos are somehow unor-
11
https://en.wikipedia.org/wiki/Daocheng_County
12
https://twitter.com/
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.