STALLA: A FRAMEWORK FOR OPEN SOURCE ANALYSIS DURING THE COVID-19 PANDEMIC
Social Networks, Recurrent Neural Networks (RNN, LSTM, BiLSTM, Weak Supervision.
The spread of social networks has resulted in an increase in the distribution of disinformation campaigns, which put national democratic stability at risk, becoming an unfavorable element for the intelligence knowledge production. In order to mitigate this bottleneck, the STALLA framework was proposed for the collection, treatment, automated labeling and analysis of information, providing greater efficiency in knowledge production. Thus, the study has as scope the Covid-19 pandemic, from data collected from short texts (tweets), in the Portuguese language, from the social network Twitter. Considering the related works, Recurrent Neural Networks (RNN) present themselves as the most suitable for textual analysis. Based on this premise, the performance of STALLA was analyzed by comparing the implementations of LSTM and BiLSTM networks, resulting in an accuracy of approximately 70\%, a value considered significant for the definition of information relevance.