Development of Kid-Friendly YouTube Access Model Using Deep Learning

Reddy, S Chakrapani; Srikanth, Nikitha; Sharvani, G. S.

doi:10.1007/978-981-15-5309-7_26

Cited by 8 publications

(6 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The authors also conducted a detailed characterization study of uploaders in terms of popularity and engagement and found close connections between child unsafe and safe content promoters. Reddy, et al [45] handled the problem of an explicit content of YouTube videos by performing text classification on YouTube comments. They processed text by performing bigram collocation and fed this to the naïve Bayes classifier for final classification.…”

Section: Related Workmentioning

confidence: 99%

A Deep Learning-Based Approach for Inappropriate Content Detection and Classification of YouTube Videos

Yousaf

Nawaz

2022

IEEE Access

View full text Add to dashboard Cite

The exponential growth of videos on the YouTube video sharing platform has attracted billions of viewers, and among them majority belongs to a young demographic. Malicious uploaders also find this platform as an opportunity to push inappropriate visual content including children-oriented videos like animated cartoons. Therefore, automatic inappropriate video content filtering is highly suggested to be integrated into social media platforms. In this study, various methods and techniques are explored for the detection and classification of child inappropriate video content. First, the proposed framework employs an ImageNet pretrained convolutional neural network (CNN) model known as EfficientNet-B7 to extract video descriptors, which are then fed to bidirectional long short-term memory (BiLSTM) network to learn effective video representations and perform multiclass video classification. An attention mechanism is also integrated after BiLSTM network to apply attention probability distribution. All models are evaluated on a manually annotated video dataset of 111,156 cartoon clips from YouTube videos that fall into three categories: safe, fantasy violence, and sexual nudity. The experimental results demonstrated that EfficientNet-BiLSTM framework (accuracy = 95.66%) performs better than attention mechanism-based EfficientNet-BiLSTM (accuracy = 95.30%). Secondly, the traditional machine learning classifiers perform relatively poor than deep learning classifiers. Overall, the combination of EfficientNet and BiLSTM with 128 hidden units in deep learning architecture yielded state-of-the-art performance in terms of f1 score (0.9267). Moreover, the performance comparison against existing state-of-the-art approaches verified the superiority of proposed framework (recall = 92.22%) in child inappropriate video content detection and classification.

show abstract

Section: Related Workmentioning

confidence: 99%

A Deep Learning-Based Approach for Inappropriate Content Detection and Classification of YouTube Videos

Yousaf

Nawaz

2022

IEEE Access

View full text Add to dashboard Cite

show abstract

“…We used Puppeteer to collect both links and emails. ab e Y T be di bi g/ i ab e ide f kid (2019) [31] Da a B a g…”

Section: Data Collection 21 Youtube Crawling and Feature Extractionmentioning

confidence: 99%

“…Another user-centered study is by Benevenuto et al [5] which comments on content pollution in video sharing platforms and provides a classification approach at separating spammers and promoters from appropriate users. Furthermore, Reddy et al [31] suggested an age detection process for underage YouTube users, supported by performing sentiment analysis on comments. In this way, the authors offer a real time protection mechanism that can be embedded in the current YouTube platform.…”

Section: Related Workmentioning

confidence: 99%

YouTubers Not madeForKids: Detecting Channels Sharing Inappropriate Videos Targeting Children

Gkolemi,

Papadopoulos,

Markatos

et al. 2022

Preprint

View full text Add to dashboard Cite

YouTube is one of the most popular social media and online video sharing platforms, and users turn to it for entertainment by consuming music videos, for educational or political purposes, advertising, etc. In the last years, hundreds of new channels have been creating and sharing videos targeting children, with themes related to animation, superhero movies, comics, etc. Unfortunately, many of these videos have been found to be inappropriate for consumption by their target audience, due to disturbing, violent, or sexual scenes.In this paper, we study YouTube channels that were found to post suitable or disturbing videos targeting kids in the past. Unfortunately, we identify a clear discrepancy between what YouTube assumes and flags as inappropriate content and channel, vs. what is found to be disturbing content and still available on the platform, targeting kids. In particular, we find that almost 60% of videos that were manually annotated and classified as disturbing by an older study in 2019 (a collection bootstrapped with Elsa and other keywords related to children videos), are still available on YouTube in mid 2021. In the meantime, 44% of channels that uploaded such disturbing videos, have yet to be suspended and their videos to be removed. For the first time in literature, we also study the "made-ForKids" flag, a new feature that YouTube introduced in the end of 2019, and compare its application to the channels that shared disturbing videos, as flagged from the previous study. Apparently, these channels are less likely to be set as "madeForKids" than those sharing suitable content. In addition, channels posting disturbing videos utilize their channel features such as keywords, description, topics, posts, etc., in a way that they appeal to kids (e.g., using game-related keywords). Finally, we use a collection of such channel and content features to train machine learning classifiers that are able to detect, at channel creation time, when a channel will be related to disturbing content uploads. These classifiers can help YouTube content moderators reduce such incidences, by pointing to potentially suspicious accounts, without analyzing actual videos, but instead only using channel characteristics.

show abstract

“…This is because such videos usually feature popular cartoon characters, like Spiderman, Mickey Mouse, etc., and include an innocent thumbnail aiming at tricking the toddlers and their custodians. In Table 3 Buzzi [62] Alshamrani [63] Reddy et al [64] Stöcker et al [65] Ishikawa et al [66] Singh et al [67] Eickhoff et al [68] Han et al [69] Alshamrani [70] YouTube Kids Tahir et al [71] -Alghowinem [72] Neumann et al [73] Other Wehrmann [74] Thierer [75] Tsirtsis et al [76] Luo et al [77] Charalambous et al [78] Ybarra et al [79] Parmaxi et al [80] Table 3.1: Studies that focus on the detection and containment of inappropriate content for children.…”

Section: Detection and Containment Of Inappropriate Content For Childrenmentioning

confidence: 99%

Characterizing Abhorrent, Misinformative, and Mistargeted Content on YouTube

Papadamou¹

2021

Preprint

View full text Add to dashboard Cite

YouTube has revolutionized the way people discover and consume video content. Although YouTube facilitates easy access to hundreds of well-produced educational, entertaining, and trustworthy news videos, abhorrent, misinformative and mistargeted content is also common.The platform is plagued by various types of inappropriate content including: 1) disturbing videos targeting young children; 2) hateful and misogynistic content; and 3) pseudoscientific and conspiratorial content. While YouTube's recommendation algorithm plays a vital role in increasing user engagement and YouTube's monetization, its role in unwittingly promoting problematic content is not entirely understood.In this thesis, we shed some light on the degree of abhorrent, misinformative, and mistargeted content on YouTube and the role of the recommendation algorithm in the discovery and dissemination of such content. Following a data-driven quantitative approach, we analyze thousands of videos posted on YouTube. Specifically, we devise various methodologies to detect problematic content, and we use them to simulate the behavior of users casually browsing YouTube to shed light on: 1) the risks of YouTube media consumption by young children; 2) the role of YouTube's recommendation algorithm in the dissemination of hateful and misogynistic content, by focusing on the Involuntary Celibates (Incels) community; and 3) user exposure to pseudoscientific misinformation on various parts of the platform and how this exposure changes based on the user's watch history.First and foremost, I am grateful to my advisor, Michael Sirivianos, for his continuous support and valuable feedback throughout my PhD journey. His support and guidance was instrumental in turning me into an independent and competent researcher. He was there to guide me when the research seemed fuzzy and disheartening. More importantly, he has shown me how to analyze an important and complex problem, and divide it in small manageable problems that can be addressed in a more practical way.

show abstract

Development of Kid-Friendly YouTube Access Model Using Deep Learning

Cited by 8 publications

References 6 publications

A Deep Learning-Based Approach for Inappropriate Content Detection and Classification of YouTube Videos

A Deep Learning-Based Approach for Inappropriate Content Detection and Classification of YouTube Videos

YouTubers Not madeForKids: Detecting Channels Sharing Inappropriate Videos Targeting Children

Characterizing Abhorrent, Misinformative, and Mistargeted Content on YouTube

Contact Info

Product

Resources

About