Identifying Citizen Interests during the COVID-19 Pandemic Using Context Change in Twitter Conversations

Events such as the recent COVID-19 pandemic tend to cause sudden shifts in people's conversations that go unnoticed by organizations at first glance. In this paper, we propose the Word Context Change metric (WCC) that detects semantic changes using a specific term during several periods by gathering users' conversations from Ecuador on Twitter (now X). We developed a machine learning model to classify tweets (now posts) based on the Oxford health policies before creating a time-Tagged corpus. Then, a temporal language representation based on word embeddings allows applying the WWC metric to determine context change relates to people's needs during the pandemic. Our experiments show that most of the emerging terms are related to Ecuador's political and health landscape during the first six months of the pandemic, while they have an emerging pattern like the search trends on Google one week ahead of the report. We conclude that our metric can anticipate text search patterns and behaviors that facilitate the identification of citizens' needs during a crisis.