TDWG2026
Home / Talks / ODIN
talk · Tuesday 22 September · 11:20–⁠11:34 · ODIN

GBIF Rule Based Annotations

John Waller · Data Quality: From Standard to Practice

Recording time 2:48:23–3:02:55Open on Vimeo ↗

Summaries and key moments are machine-produced. Check the recording for the speaker’s exact words.

The short versionSimple, scoped negative-range rules let non-experts flag suspicious GBIF occurrences persistently, including future records.

Overview

What this was about

John Waller presents experimental work at GBIF on rule-based annotations that capture 'rules in people's heads' about suspicious occurrences, e.g. 'amphibians in Antarctica are suspicious'. Rules are polygons plus taxon and optional scoping (dataset key, basis of record, fossils excluded) and apply to existing and future records. Examples include an elephant shrew record in Madagascar and axolotls outside Mexico; he argues that marking where a species is certainly not found (an 'anti-distribution') is much easier for non-experts than defining native ranges. A demo Download Annotator and an R package, gbifrules, apply rules to downloads.

Why it matters. It offers a lightweight, community-driven way to capture expert knowledge about data errors that survives unstable occurrence identifiers.

Key ideas

In the room

  • Rules flag all current and future matching records (taxon within polygon) as suspicious.
  • Anti-distributions (where a species will not occur) are easier to define than native ranges and need little expertise.
  • Because GBIF occurrence IDs are not fully stable, single records are targeted by scoping rules to a small box plus dataset key and basis of record.
  • Complex rules allow exceptions, e.g. no non-human primates in Cuba except fossils.
  • Tools: an experimental web UI for making rules, a browser-based Download Annotator demo, and the gbifrules R package (GitHub, not CRAN).
  • Main challenge: recruiting rule makers; the feature is not yet visible on gbif.org.
Jump in

Notable moments

In their words

Transcript

Automatically generated captions can contain mistakes, especially in names and technical terms. Times are relative to the room recording.

Read the transcript ↓
Loading transcript…
Keep exploring

Related talks.

See the map ↗
talk · 22 Sept · 12:00–⁠12:15 · SAL B

Geonomia: Metadata and Community for Georeferencing

Clustering specimens into collecting trips makes georeferencing and AI enrichment more efficient and should be done as a community effort.

Nicky Nicolson
talk · 22 Sept · 14:00–⁠14:15 · SAL B

Open Digital Curation and Round-Tripping Biodiversity Data Enhancements Across Collection Infrastructures: Six Years and 50 Million Links in Bionomia

Round-tripping is much harder than expected, and data managers must fully populate identifiedBy so taxonomists' expertise is credited.

David Shorthouse
talk · 22 Sept · 15:00–⁠15:15 · SAL B

Closing Gaps in Specimen Metadata: A Collaborative Annotation and Feedback Platform for Natural History Collections

A national collaborative annotation platform with AI-assisted transcription can close metadata gaps across many Taiwanese collections.

Szu-Hsien Lee
Programme abstract

GBIF Rule Based Annotations

John Waller

Biodiversity data quality is commonly run through automated tests and standardized assertions, yet many data quality issues are not captured by existing pipelines.

We present rule-based annotations as a practical approach for translating informal expert knowledge into structured, reusable data quality rules. Simple statements such as “there are no dragonflies in Antarctica” can be formalized as rules linking taxa, geography, and controlled vocabularies (e.g. suspicious, native, introduced), and applied to both current and future occurrence records.

Unlike traditional cleaning approaches or one-off feedback to data publishers, rule-based annotations are persistent, shareable, and extensible. Rules can incorporate additional constraints (e.g. basis of record, time, dataset) and are managed through an experimental community-driven system. This enables a flexible, opinion-based layer of data quality assessment that complements existing standards and automated checks .

We demonstrate the approach using a case study on Odonata, where expert-defined rules identify a small number of outliers not captured by standard cleaning pipelines, yet with disproportionate effects on downstream species distribution models.

Just as filtering rules in an email inbox serve to filter and sort incoming emails, rule based annotations are meant to be future looking. Once a rule is created it is saved to a database, and is accessible to all GBIF users. Users can leave comments and upvote or downvote based on their perception of the general usefulness of the rule.

Submitted to the conference programme before the talk, so the talk itself may differ.