2019-09-15, 12:00–12:30, Ferrier Hall
Uncovering Twitter troll armies by monitoring and analyzing millions of tweets using Python to identify suspicious entities that intended to skew online conversations and spread misinformation.
Using Twitter Streaming API we collected and analysed millions of tweets to identify suspicious entities like bots and their impact on online conversations over a period of 3 months. We aim to understand the evolution, affiliation and participation of trolls directly or indirectly to skew public opinion and spread misinformation.
In our talk, based on the analysis, we would like to share the following:
* Identification of trending topics along with for-and-against Hashtags surrounding a given topic.
* Classification of users based on the likelihood of suspicious activity
* Monitoring of above users’ accounts to understand their individual and collective impact and implication on steering of the online conversation
* Tools used to manage, collect and analyze the dataset:
* Bokeh etc.
Based on our leanings we have created a framework with Python ecosystem that monitors tweets in real-time and comes with the following features: * An Interface that easily configures keywords/hashtags to be monitored * Classifies users in real-time and monitors their activity * Identifies popular hashtags/keywords and feeds them back for monitoring their usage. * Dashboard that displays the following insights: * Number of unique tweets * Percentage tweets from suspicious users. * Volume of tweets from suspicious users, contributing to popular hashtags. * Flagging the compromised hashtags. * Top Users * Top Hashtags * Top Keyword
Takeaways for the audience * Creating awareness about open twitter data and its research possibilities. * A quick overview of how bots are all-pervasive and impacting online conversations. * Provide a self-service tool built entirely in Python eco-system that empowers users and researchers to verify Twitter conversations’ authenticity.