Accepted Poster
Poster Short Abstract
The iNaturalist Tagger lets citizens describe biodiversity images with tags. User inputs are compared with large language model outputs to study how invasive and endangered species are described. Evaluation includes completion rate, skip rate, and tag diversity.
Poster Abstract
Images of organisms in natural settings are an increasingly important source of information for biodiversity records. However, beyond species names and occurrence records, richly descriptive tags can capture habitat context, behavior, visual traits, and human perceptions, information that is often not included in traditional metadata. Semantic tagging thus enhances data integration, searchability, and downstream applications such as ecological modelling, automated classification, and conservation assessments. The iNaturalist Tagger is a citizen-centered web app that allows citizens to add image description into already published records in iNaturalist. Built in R Shiny, the GUI asks users to annotate randomized animal images from the Helsinki Metropolitan Area HMA-LTSER (Finland) published on iNaturalist. In each image, the user is asked to provide short tags about the image (describing the animal, the habitat where it is, and/or the user feelings towards the image and/or the animal in question). The species vernacular name (common name) and contextual badges (species conservation status and invasiveness in the study area) appear below each image providing additional information to users. The tags provided by the users will be used for comparison with large language model (LLM - chatGPT) outputs to understand the variation in semantic descriptions of invasive and endangered species between humans and LLM. Planned evaluations include completion rate, skip rate, tag diversity. Through a user-friendly interface and lightweight infrastructure, iNaturalist Tagger offers a reusable template for citizen science that improves semantic annotation of online biodiversity images.
Poster Session