| Popis: |
Social media is increasingly scraped to extract information on various topics. Within EU-funded projects, text mining is being applied to gain insights into public perceptions on climate change, citizen science activities, and on the relationship between climate, health, and ecosystems. For this purpose, natural language processing tools were used to automatically pre-process, clean, and extract meaningful segments from the text posted on social networks. The proposed text mining tools are building on the open-source pre-training method for natural language understanding (BERT) and use language models pre-trained on large amount of generic and domain specific texts. With this method, domain-specific meaningful phrases, named entities (places, organizations, etc.), and overall post sentiment are extracted to gain an understanding of the public perceptions: in the H2020 I-CHANGE project, we analyzed posts from Twitter to gain an understanding of current trends around climate change and citizen science but also to summarize people’s perceptions of risks derived from climate change and hydrometeorological hazards. While in the HEA TRIGGER project, text mining is used to detect and assess the human perception of health impacts from climate stressors. |
