Seungju Han, Beomsu Kim, Jin Yong Yoo, Seokjun Seo, Sangbum Kim, Enkhbayar Erdenee, Buru Chang. Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2022.
Social media platforms struggle to protect users from harmful content through content moderation. These platforms have recently leveraged machine learning models to cope with the vast amount of user-generated content daily. Since moderation policies vary depending on countries and types of products, it is common to train and deploy the models per policy. However, this approach is highly inefficient, especially when the policies change, requiring dataset re-labeling and model re-training on the shifted data distribution. To alleviate this cost inefficiency, social media platforms often employ third-party content moderation services that provide prediction scores of multiple subtasks, such as predicting the existence of underage personnel, rude gestures, or weapons, instead of directly providing final moderation decisions. However, making a reliable automated moderation decision from the prediction scores of the multiple subtasks for a specific target policy has not been widely explored yet. In this study, we formulate real-world scenarios of content moderation and introduce a simple yet effective threshold optimization method that searches the optimal thresholds of the multiple subtasks to make a reliable moderation decision in a cost-effective way. Extensive experiments demonstrate that our approach shows better performance in content moderation compared to existing threshold optimization methods and heuristics.
Videos are dynamic and multi-modal compared to other types of content, making automatic filtering difficult, which is why content moderators play a crucial role. However, video content moderators are exposed to more profound emotional labor because videos contain rich visual information, sometimes including even harmful content, such as violent or terrifying scenes. In this work, we explore the effect of six intervention techniques on alleviating negative emotions during video content moderation tasks. We conducted one online crowdsourcing experiment and two controlled user studies to find out that (i) interleaving with positive videos or (ii) cartoonization could significantly reduce negative emotions in the moderators. Participants reported that the advantages of these approaches are in helping reduce negative emotions at the time of moderation while existing approaches focus on post-task activities (e.g., relaxation, talking with others, or getting a hobby). We discuss the applicability of our findings to broader tasks, including improvement in intervention techniques.
Normalization techniques are known to provide faster model convergence and good generalization performance, having achieved great success in computer vision and natural language processing. Recently, several deep neural network-based click-through rate (CTR) prediction models have applied such normalization techniques to their deep network components to make model training stable. However, we observe that applying existing normalization techniques (e.g. Batch Normalization and Layer Normalization) to feature embedding of the models leads to the degradation of model performance. In this study, we conjecture that existing normalization techniques can easily ignore the significance of each feature embedding, leading to sub-optimal performance. To support our claim, we theoretically show that existing normalization techniques tend to equalize the norm of individual feature embedding. To overcome this limitation, we propose a theory-inspired normalization technique, called Embedding Normalization, which not only makes model training stable but also improves the performance of CTR prediction models by preserving the significance of each feature embedding. Through extensive experiments on various real-world CTR prediction datasets, we show that our proposed normalization technique leads to faster model convergence and achieves better or comparable performance than other normalization techniques. Especially, our Embedding Normalization is effective in not only deep neural network-based CTR prediction models but also shallow CTR prediction models that do not utilize deep neural network components.