Social-Media-Disinformation-Network-BERT | Social_Media_Disinformation_Network Twitter

Social-Media-Disinformation-Network-BERT is a Jupyter Notebook library. Social-Media-Disinformation-Network-BERT has no bugs, it has no vulnerabilities, it has a Strong Copyleft License and it has low support. You can download it from GitHub.

# Social_Media_Disinformation_Network Twitter is a social networking platform where many political thoughts and views are exchanged between users. Some of the users are, in fact, nation state actors – individuals having close links to the military, intelligence or state control apparatus of their country – who share fake news to engage in espionage, propaganda or disinformation campaigns. Twitter has already identified many of these accounts and banned them from Twitter for violating Twitter policies. Our main goal is to build a classification Natural Language Processing (NLP) model by learning disinformation and fake news patterns from tweets and to classify them either as “Disinformation” or “Others.” This study makes use of state-linked information operations (“IO”) data published by Twitter in June 2020 covering operations attributed to Russia and Turkey. We narrowed our focus to the Turkish and Russian tweets which were involved in a range of manipulative and coordinated activities spreading geopolitical narratives favorable to their respective political parties in Turkey. For our classification model we also incorporated Twitter live stream data from the Twitter archives for the same time period. Using SQL queries, we isolated the 8,392 banned Turkish & Russian accounts from the archived live stream data to create our “Others” category data. Using a Bidirectional Encoder Representation from Transformers (BERT) model, with the “Turkey” & “Russia” information operations and “Others” live stream archive category data for training, we tested this model against archived Twitter tweets for the month time period following the time period of the training data. Our model predicted 43,568 tweets as “Turkey” disinformation out of 411,095 tweets with an accuracy of 89.4%. For the same time period Twitter banned only 26,259 disinformation tweets. Based on our prediction model it appears that Twitter may still be missing 17,309 information operations tweets for that time period, Similarly our model predicted 20,826 tweets as “Russia” disinformation out of 114,416 tweets with an accuracy of 81.79%.