Abstract of presentation:
Fake news, fake reviews, hating or trolling are the most prominent examples of online antisocial behavior which is currently present across the whole web. Growing negative consequences of such negative behavior have recently elicited many efforts, which are aimed at its characterization, detection and mitigation.
At first in our talk, we will shortly describe how the state-of-the-art data-driven techniques (especially AI, ML, data mining, NLP) can be applied in fighting antisocial behavior. We will highlight the most challenging open problems, such as a lack of suitable rich datasets or useful end-user services as well as bad interoperability of detection methods.
Secondly, we will introduce our solution how to overcome the stated open problems. We are working on a novel, universal and easily extensible platform named Monant. Our platform provides features for web monitoring, interfaces for AI classification and detection methods as well as for end-ser services (e.g. how to train users to detect fake news).
In more details, we will discuss its architecture and current state of implementations using many Python libraries for web crawling/parsing, scheduling extractions, API communication between main platform modules, etc.