Volume 12 - Issue 3
Classifications of Restricted Web Streaming Contents based on Convolutional Neural Network and Long Short-term Memory (CNN-LSTM)
- Jaewon Choi
Department of Business Administration Soonchunhyang University, Asan-si, 31538, South Korea
jaewonchoi@sch.ac.kr
- Xiuping Zhang
Department of Business Administration Soonchunhyang University, Asan-si, 31538, South Korea
zhangxiuping@sch.ac.kr
Keywords: Content classification, Convolutional neural network, Long short-term memory
Abstract
The development of social media is beneficial for users to quickly access various types of information
online. However, this can be a risky for teenagers under the age of 18 years because they
may become exposed to information that is unsuitable for them. Some social media platforms have
established age-restricted content policies to prevent teenagers from being exposed to bad information.
However, unsuitable content that has not been marked as age-restricted still exists online as a
result of the enormous volume of information provided on the Internet and the inability to identify
it immediately, among other factors. It is important to classify restricted and unrestricted content to
protect teenagers’ online safety because teenagers are more likely to be negatively affected by biased
and harmful content than adults are. We suggest a strategy for classifying restricted and unrestricted
content in this study by examining content comments. We collected and cleaned comments obtained
from two datasets (each containing restricted and unrestricted content comments, respectively) from
YouTube. Word2vec was used to display comments as vectors, and the classifier was established using
convolutional neural network and long short-term memory. Through our findings, we hope make
the social media environment more secure to protect the physical and mental health of teenagers.