Due to the increasing number of new and reemerging pests resulting from intensification, globalisation and climate change, monitoring of plant health is crucial. In this context, outbreak detection in digital media could be useful for improving plant disease surveillance. But manually extracting relevant information from unofficial sources is time-consuming. The Platform for Automated extraction of Disease Information from the web (PADI-web) has been developed initially for animal health surveillance, and recently for plant disease surveillance. In order to identify relevant news and information with this new PADIweb instance dedicated to plant health, machine learning approaches and language models (RoBERTa) have been integrated for monitoring plant diseases. This paper presents the PADI-web algorithms and visualisations implemented for specific case studies (i.e. Xylella fastidiosa and Fusarium Oxysporum Tropical) using text-mining approaches tuned on the plant disease domain.