Signals from the deep web - Making sense of ephemeral and anonymous content by Ylli Prifti

With recent development in Computational Social Science we see an increase in research projects using mainstream social media APIs for access to data. It has been shown that APIs can be biased, less representative and suffer from unintended use limitations. We propose an alternative to use of APIs that is based on data scraping and have used oXpath, part of the wider Vadalog framework to bridge Machine Learning and Reasoning. In this session we will propose a scalable architecture for capturing data from the deep web, using an implementation for a Demo. We will show advantages over conventional API usage, and will use Topic Modelling and Sentiment Analysis on captured data to try bring insight on how a particular community might be manipulated by use of bots and other means.