Natural Language Processing Meets Computational Social Science

The ongoing global discussions on environmental issues such as air and water pollution, climate change, and reduced species diversity have highlighted the need for new methods and tools to study and understand the complexity of these relevant issues. This work uses modern text processing methods to analyze the discourse of social media related to ecology and environmental aspects. The study is based on tag-sampled data from Russian news agency “Ria Novosti” over the past few years. The main approaches combine topic modelling approach: latent semantic analysis, latent Dirichlet allocation, Embedded topic model and text classification (for mining texts related to ecology). These approaches allow to collect a representative body of environmental texts, identify keywords and topics, and trace hidden semantic topics and follow them through time. The results of the study will allow us to assess the current environmental situation through the prism of news sources, as well as to track the dynamics over the past few years.