Rumor Detection using Deep Learning on Twitter
Important information during disaster situation can be reached billions and billions of people via the social medias like Twitter, Facebook,Instagram etc. without too much effort and time but to justify these news is still a great issue.Rumors always have great impact to the persons or to the society.Especially the spread of the false rumors on the social medias during disaster or emergencies can be the main cause of huge loss of time, resources and even also the life of the humans . It is always very hard to di erentiate the rumors to non-rumors during disaster situation. The main objective of this research project is to familiarize and acquire the deep understanding of the state of the art research on the modern deep learning methods and how it can be used on the social media platform like Twitter .to classify information rumors or non-rumors.We investigate and implement the popular Recurrent Neural Network (RNN) and the variants of RNN like Long Short-Term Memory (LSTM) and Gated Recurrent Unit(GRU) in order to classify the tweets as rumor and non-rumors. All of our models learn to classify in the supervised environment.For this project, we used the PHEME[1] dataset which contains in total 5,802 rumor and non-rumor tweets related to the real world events like the Ferguson unrest, Ottawa shooting, Charlie Hebdo, Germanwings crash and Sydney siege.All of our deep learning models achieved to classify rumors to non-rumors with an accuracy of more than 80% and the model GatedRecurrent Unit(GRU) achieved with an accuracy of 89%.