Angiosperm functional traits extracted from botanical descriptions

This dataset provides a large-scale compilation of plant functional traits automatically extracted from published taxonomic descriptions of flowering plants. The taxonomic backbone comprises all accepted angiosperm species names sourced from World Flora Online (WFO). Each accepted species was systematically searched across multiple textual sources, including the Biodiversity Heritage Library (BHL; OCR-derived biodiversity literature), eFloras (including Flora of North America, Flora of China, and related online floras), and the Ecological Flora Database (Ecoflora). Where botanical descriptions were identified, natural language processing methods were used to extract structured trait information.

Trait extraction was implemented using the [ADEPT](https://github.com/NaturalHistoryMuseum/ADEPT) pipeline, which combines machine-learning classification of botanical descriptions, ontology-driven terminology matching, named-entity recognition for numeric traits, syntactic parsing, and rule-based semantic interpretation to associate traits with plant structures and standardise measurements.

The dataset includes trait information for 59,369 angiosperm species, comprising 888,328 extracted trait values spanning vegetative, reproductive, and ecological characteristics such as growth form, plant height, leaf morphology, floral traits, fruit and seed traits, dispersal attributes, and cytological information where available. Trait coverage varies among taxa and trait types, reflecting differences in descriptive detail, taxonomic practice, and source availability.

Separate files are provided for each source dataset alongside a combined dataset, enabling users to assess provenance and tailor analyses. This resource supports large-scale studies of plant functional diversity, ecological strategy, conservation risk, and biodiversity modelling by mobilising trait information historically embedded in taxonomic literature.