We present a holistic approach to building a robust and useful natural language classification system for real-world content moderation. The success of such a system relies on a chain of carefully designed and executed steps, including the design of content taxonomies and labeling instructions, data quality control, an active learning pipeline to capture rare events, and a variety of methods to make the model robust and to avoid overfitting. Our moderation system is trained to detect a broad set of categories of undesired content, including sexual content, hateful content, violence, self-harm, and harassment. This approach generalizes to a wide range of different content taxonomies and can be used to create high-quality content classifiers that outperform off-the-shelf models.

The CVP Cyberspace Visualization Project addresses interactive overview visualizations of Web spaces for enhancing orientation, navigation and communication when browsing or searching the WWW. This paper presents a collection of requirements which such visualizations should fulfil. An algorithm is described generating recognizable, hierarchically structured views of interconnected Web objects. The CVP is outlined as improving current tools for Web navigation and communication. It combines 3D overviews with the mutual representation of users, providing first steps towards the Web browsers of the next generation.