4 1. INTRODUCTION
ganized. Different from the traditional long videos that can be well-structured into a specific
video genre from “Crime,” “Documentary,” “Romance,” to “War,” like the video organization
in YouTube, micro-video, as a new medium, does not have a matured taxonomy to follow. Also,
it is inappropriate to apply the long video taxonomy to micro-videos due to their different em-
phases. In particular, micro-videos record the actual things in life, whereas the long ones cover
a broader range of things like marvelous performance.
ankfully, a micro-video is frequently recorded at a specific place within one shot, and
hence micro-video services are able to encourage users to manually label the micro-videos with
GPS-suggested venue information [49], such as “Orchard ION Singapore.” Each venue belongs
to a venue category, such as “shopping mall,” based on the Foursquare API
13
and the venue
categories are organized into a tree-like taxonomy.
14
We show part of the tree structure in the
Figure 1.2. From the figure, we find that: (1) Foursquare organizes the venue categories into a
four-layer hierarchical tree structure; and (2) the top layer of the tree contains ten non-leaf nodes
(coarse venue categories). We aim to organize micro-videos by categorizing them into the leaf
nodes of this tree. It will benefit multifaceted aspects: (1) footprint recordings—it facilitates
users to vividly archive where they were and what they did; (2) personalized applications—
such people-centric location data enables precise personalized services, such as suggesting local
restaurants, alerting regional weather, and spreading business information to nearby customers;
and (3) other location-based services. Location information is helpful for the inference of users’
interests, the improvement of activity prediction, and the simplification of landmark-oriented
video search.
Despite its significance, users of micro-video platforms have been slow to adopt this
geospatial feature: in a random sample over 2 million Vine videos, we found that only 1:22% of
the videos are associated with venue information. It is thus highly desirable to infer the missing
geographic cues.
1.2.3 MICRO-VIDEO ROUTING
Along with the popularity of micro-videos, users are frequently overwhelmed by their unin-
terested ones. It becomes increasingly difficult and expensive for users to locate their desired
micro-videos from the vast candidates. is is due to the following reasons: (1) existing recom-
mendation systems developed for various communities cannot be straightforwardly applied to
route micro-videos, since users in micro-video platforms have their unique characteristics, such
as the complex interactions between users and micro-videos; and (2) in micro-video platforms,
users follow their interested topics and the platforms usually recommend users with the micro-
videos falling in the range of the followed topics. However, users’ interests are dynamic and
evolve over time. In light of this, it is crucial to build a personalized recommendation system to
intelligently route micro-videos to the target users, which will strengthen customer stickiness.
13
https://github.com/mLewisLogic/foursquare
14
https://developer.foursquare.com/categorytree
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.