146 7. RESEARCH FRONTIERS
for the thumbnail selection. Based on the constructed dataset, the publisher’s preferences on dif-
ferent frames are captured by analyzing the multimodal content information of micro-video and
the publisher’s profile. en, in order to calculate the popularity of each frame, the distribution
of users’ interests on the platform can be modeled via exploring the comments of published
micro-videos. Ultimately, an attractive thumbnail can be obtained by combining the publisher’s
preference and popularity with the quality and representativeness.
7.4 SEMANTIC ONTOLOGY CONSTRUCTION
With the flourish of plentiful micro-video communities and exponentially growing number of
micro-videos, it is necessary to help users quickly find the micro-videos they are interested. As is
known to all, the classification to the labels of micro-video is an important aspect to save the user
overwhelmed in large amounts of micro-videos. However, the labels of micro-videos are usually
uploaded by users, whose laziness caused the existing micro-video only has the automatically
acquired geographical label, but lack that of the semantic ontology, which plays a pivotal rule in
the content-based micro-video search. In fact, labels of an instance is generally correlated with
certain structure, such as the category-level hierarchical structure between cooking methods and
cooking dishes, and so is the semantic ontology labels of the micro-video. Actually, the structured
hierarchical labels are able to exclude several noise when searching the desired micro-videos. For
example, in Figure
7.5, the cooking videos of same cooking method are gathered in one branch,
and all the cooking videos are clustered in a tree, which is convenient for the user who is seeking
for the micro-video about the cooking method instead of the review or exhibition of the food.
In light of this, the hierarchical semantic ontology construction and classification based on it of
the micro-video get our attention.
Bread … ……
… …
Biscuits
Bake Fry
Cook
Stew
Fried Chicken French Fries Bortsch
Tomato
Beef Brisket
Figure 7.5: Example of hierarchical cooking videos.
Although several methods have been dedicated to the onefold semantic ontology clas-
sification of micro-video, the hierarchical semantic ontology classification remains largely un-
tapped. Limited efforts have been devoted to the hierarchical semantic ontology classification
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.