Course project for Text Analysis and Retrieval at Faculty of Electrical Engineering and Computing, University of Zagreb
The tasks of rumour veracity and stance classification have captured the interest of researchers with the rise of social media and its consumption for relevant information spreading. Traditional approaches of feature engineering have been challenged by neural language models. This paper provides a broad overview of the performance of different feature setups across several classifier models and show- cases that this approach is still relevant for the stance classification task. Furthermore, we also investigate how including the simplest form of discussion environment, the discussion source tweet, influences the performance of a simple neural model classifier.