CMTA: A framework for Multilingual COVID-19 Tweet AnalysisDownload PDF

Sep 03, 2020 (edited Sep 03, 2020)EMNLP 2020 Workshop NLP-COVID SubmissionReaders: Everyone
  • Keywords: COVID-19, BERT, Multilingual, Twitter, Social Computing, NLP
  • TL;DR: A emotion and misinformation based multilingual analysis of COVID-19 tweets.
  • Abstract: In the current scenario of COVID-19 pandemic when people around the world are facing restrictions in their daily life activities, there is a rapid growth in the number of people relying on the internet. Billions of people are using social media platforms such as Twitter for sharing COVID-19 related news, information and thoughts which reflect their perception and opinion about the pandemic. Analysis of tweets for identifying misinformation and emotion analysis can generate valuable insights. We present CMTA: a multilingual COVID-19 related tweet analysis model. In our work, we propose a deep learning model for multilingual tweet misinformation and emotion detection and classification. CMTS uses multilingual BERT for extracting features from multilingual textual data, which is then categorised into specific emotion and misinformation class. Classification is done by a Dense-CNN model trained on tweets manually annotated into emotion(8 classes) and misinformation( 5 classes). We also present an analysis of multilingual tweets from April to June month showing the distribution of public emotion and misinformation spread across different languages.
6 Replies