Summaries and key moments are machine-produced. Check the recording for the speaker’s exact words.
The short versionTaxonWorks and DwC-DP have converged conceptually, but exchange between rich models will always be lossy, so mappings should be recorded explicitly.
↶Go back to 2025Data Package pilot · 1 connection↓
Revisit Living Data 2025 alongside this talk. These connections were checked against captions; related examples may come from different teams rather than a direct follow-up.
Related example
Near-lossless mapping versus lossy exchange
2025 Living Data
Sam Leeflang reports that openDS and DwC-DP are closely aligned and little data is lost in transformation.
Seamless Integration: How DiSSCo's openDS Facilitates Data Interoperability with the Darwin Core Data Package
implementation is Yeah. So what are the lessons that we have learned from this uh implementation? Well, the interoperability is is is good. We we don't lose a lot of data in the transformation. Um we do need to do some triggery uh in the sense that yeah, we we have this document approach. Um and a lot of the relationships that we have are actually implicit in it because they're nested into in this in this bigger object. we don't need an specific identifiers for them. Um, so what we did need to do was create a lot of identifiers which we do through an MD MD5 hash of the object and then uh we needed to duplicate all the information where uh where we use temporary tables. So we we hash those and and use those uh
2026 This talk
Matthew Yoder says partial alignment is normal and exchange from TaxonWorks to DwC-DP is lossy.
Aligning An Evolving Global Data Model To Software With Global Ambitions: GBIF x TaxonWorks
agents. And I'll just zoom ahead to my conclusion here. It was a wonderful exercise to compare against that DP package, and I think it's becoming much easier to align broader conceptual frameworks. So that's great. One of the things that's coming out, to the chagrin of my German colleague's conversations, is that there are certain patterns that are coming out. Partial alignment is the norm. Lossy exchange. If I go from TaxonWorks to DP, I'm always going to be lossy. And for argument's sake, I call this a law because it happens everywhere. It's not a law, right? But how does that help us think about, and is it a barrier to describing life at scale if we always have to go lossy to lossy? Will SSM be adequate for schema exchange? I hope so. It should be cool. We see schema convergence. It's very clear to me that we have schema convergence
Mapping quality depends on the source model and use case; this is a useful contrast between two projects.
About this connection
Different systems are being compared. This is neither a regression from 2025 to 2026 nor evidence that either speaker contradicts their own earlier claims.
This pilot covers one 2025 symposium and selected 2026 moments. Automatic captions can contain mistakes; audio has not been checked.
Overview
What this was about
Matt Yoder compared TaxonWorks' API-first data model with the new Darwin Core Data Package (DP). He found the conceptual overlap 'uncanny' after years of independent development. DP's many media, identifier, protocol and reference link tables collapse into single polymorphic annotation tables in TaxonWorks. TaxonWorks lacks a primary Survey concept and full nucleotide analysis, while DP lacks precise anatomy links and physical container hierarchies. He proposed recording mappings (broader/narrower) in SSSOM, merging them in RDF into a 'data mapping of all data mappings', and exposing /vocab endpoints. His observations: partial alignment and lossy exchange are the norm, both models have competing 'fuzzy buckets', and DP's role as an exchange format is unclear.
Why it matters. It is a concrete software–standard alignment exercise with ideas (SSSOM mappings, /vocab endpoints) that could help many CMSs map to DwC-DP. It also flags practical issues like duplicate extension buckets.
Key ideas
In the room
TaxonWorks serves taxonomists first and now about 10–11 large collections; it is API-first and set up for human and agentic contributions.
It already offers Darwin Core import/export, a built-in IPT-like GBIF endpoint, EML persistence and lossless DwC import; an Orthoptera project round-tripped over a million GBIF records.
Conceptual overlap with DwC-DP is remarkable; TaxonWorks models one image depicting many concepts via a single annotation pattern.
Gaps in TaxonWorks: Survey as a primary concept, full nucleotide analysis; gaps in DP: precise anatomy, physical containers, loans, stronger geo-reference primacy.
Mappings can be expressed as broader/narrower relations and persisted in SSSOM, merged in RDF.
Both models have two 'buckets' for fuzzy data (resource relationship vs assertions; observations), which will confuse users.
Needs identified: equipment and configuration, experiments, genomes, more strongly typed agents; comments used as definitions cause confusion.
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.
