Identification of rumours in social networks

In a world of 7.6 billion population, it is quite challenging to collect information from each and every corner of the world about every person or about every event/incident. With the rapid growth of technology, people have a wide range of options to communicate across the world. But, the easiest as well as the most trending platform to share information, images and videos is the emerging ‘social media platforms’. Information shared over social media is extraordinarily rapid, faster than traditional media such as TVs and newspapers and can reach everyone within few seconds just by the click of a button. But how far such information is reliable and authentic, is the biggest concern. Rapid spread of “any” kind of information can also lead to the spread of many misleading or false information. As such information reaches people without any prior verification for authenticity, there are high chances of chaos and anxiety among people, unnecessary panic, wastage of time, and even loss of life and property especially during disaster. So, the need to establish efficient methods to detect these rumours and study their behaviour of dissemination i.e. the way they ‘spread’ all over the social media, and take necessary actions to combat them, has become not only important but also quite challenging. In this thesis, several techniques used to detect rumours have been discussed briefly, and out of these, predicting the ‘rumour’ and studying its variation with time is the primary focus. This is shown by the help of ‘Time Series Variation’ of detecting rumours, using graphical transformation approach and graph edit distance, which tells about the possibility of a certain information being a ‘rumour’, by analysing its dissemination over time. Once, the possibility of a rumour is detected, it is compared or checked with reliable news sources to finally approve of it being a rumour or not.Master of Science (Signal Processing