A new collaborative study published in Nature Plants introduces an innovative artificial intelligence approach to understanding biodiversity. The paper, “Learning the syntax of plant assemblages”, presents Pl@ntBERT, a language-based AI model that learns the “syntax” of plant communities much like large language models learn the structure of human language.

Pl@ntBERT was trained on extensive European vegetation data and can predict missing plant species in ecological surveys while accurately identifying habitat types. By modelling the hidden relationships among species, including shared environmental preferences, indirect interactions, and community organisation, the model substantially improves traditional classification methods used in biodiversity mapping, conservation, and restoration.

Just as the arrangement of words in a sentence determines its meaning, the abundance and ordering of plant species in a community reflect the underlying ecological “grammar” that shapes ecosystems. Understanding this syntax provides new insights into how species interact, coexist, and respond to environmental change.

This work demonstrates how AI, inspired by natural language processing, can transform ecological monitoring and provide powerful new tools for safeguarding biodiversity in an era of rapid environmental change.

The research described in this paper was funded by the European Commission through the MAMBO and GUARDEN projects.