University of Glasgow

Roderic Page

These are simply notes to myself about taxonomic classifications in
Wikidata.  
  
Classifications in Wikidata can be complex and are often not trees. For
example, if we trace the parents of the frog family Leptodactylidae back
we get a graph like this:  
  

::: {.separator style="clear: both; text-align: center;"}
[![](https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgrVA0rhFnWhs7quWHIXjbE6vsQs-gWbfLDu9NL8wrFWCUfXSLmnKzQJRlaqoNujlXccyZnKYUNWR4jEG8k8HiV_6ZR4D7harHHJSgtydn5RVZF6IONOfud_i88cqdSppoMED5O/s640/Leptodactylus-lineage_dag.jpeg){border="0"
original-height="1600" original-width="335" height="640"
width="134"}](https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgrVA0rhFnWhs7quWHIXjbE6vsQs-gWbfLDu9NL8wrFWCUfXSLmnKzQJRlaqoNujlXccyZnKYUNWR4jEG8k8HiV_6ZR4D7harHHJSgtydn5RVZF6IONOfud_i88cqdSppoMED5O/s1600/Leptodactylus-lineage_dag.jpeg){imageanchor="1"
style="margin-left: 1em; margin-right: 1em;"}
:::

  
Each oval represents a taxon in Wikidata, and each arrow connects a
taxon to its parent(s) in Wikidata.  Likewise, if we do the same for the
albatross genus *Diomedea* we get a similarly complex diagram:  

::: {.separator style="clear: both; text-align: center;"}
[![](https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgW4Kkrq9jXIe4f8y2nSR72AnD8ZQc3hd0wKv9Z-vquu8cxfxMFl-5c0a-GSERsTl6tqQ1-0Y9kPySQ90AlwVqsCPMGGR4tyotJSaGYv-9lizOjnKrKmhNeLep-_9sgf3b2E6T2/s1600/Diomedea-lineage_dag.jpeg){border="0"
original-height="1600"
original-width="135"}](https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgW4Kkrq9jXIe4f8y2nSR72AnD8ZQc3hd0wKv9Z-vquu8cxfxMFl-5c0a-GSERsTl6tqQ1-0Y9kPySQ90AlwVqsCPMGGR4tyotJSaGYv-9lizOjnKrKmhNeLep-_9sgf3b2E6T2/s1600/Diomedea-lineage_dag.jpeg){imageanchor="1"
style="margin-left: 1em; margin-right: 1em;"}
:::

  
  
The presence of multiple classifications likely reflects several
factors. If you deal with just extant species you are likely to have
fairly shallow classifications, for example, the kingdom, phylum, class,
order, family, genus ranks used by GBIF. may be enough. Some taxonomic
groups may routinely use ranks such as subfamily, and in well-studied
groups there may be additional taxa based on phylogenetic research
(e.g., the [RTA clade](https://en.wikipedia.org/wiki/RTA_clade) in
spiders). And of course, different Wikidata editors may favour different
classifications.  
  
Anecdotally (certainly for vertebrates), many of the additional levels
in the classifications in Wikidata come from fossil taxa. In the case of
birds, extant Aves (birds) are a fairly isolated group in the tree of
life, but as we go down the tree towards their common ancestor with the
crocodilians we encounter dinosaurs and other taxa. So if you are a
palaeontologist the jump from, say Aves to Tetrapoda skips over a fairly
significant part of the tree!  
  
Faced with this complexity, how do we display a Wikidata classification
in a simple way? One approach may be to display only a classification
from a particular source, for example Mammal Species of the World. This
requires that Wikidata has that classification, and enough information
for you to extract it by a SPARQL query (for example if each node in the
classification that is in MSW has a reference to MSW attached to that
node).  
  
Another approach is to extract a simplified classification from the sort
of graphs shown above. Technically, these graphs are DAGs ([Directed
acyclic graphs](https://en.wikipedia.org/wiki/Directed_acyclic_graph)).
An obvious way to simplify a DAG is to find the shortest path in that
DAG. For example, the path (Eukaryota, Animalia, Bilateria,
Deuterostomia, Chordata, Olfactores, Gnathostomata, Tetrapoda, Amphibia,
Anura, Leptodactylidae) is a path through the DAG shown above. Shortest
paths are reasonably easy to find once you have a [topological
sorting](https://en.wikipedia.org/wiki/Topological_sorting#Application_to_shortest_path_finding)
of the graph (see e.g. [Shortest Path in Directed Acyclic
Graph](https://www.geeksforgeeks.org/shortest-path-for-directed-acyclic-graphs/)).
At the moment this looks the best bet for displaying classifications
from Wikidata.  
  

#### Preferred classifications

In some cases the classification in Wikidata is complicated, but this
complexity isn't reflected in SPARQL results because parts of that
classification have different "ranks". For example, for the plant order
[Fagales](https://www.wikidata.org/wiki/Q21881) there are currently
seven parents:  

- fabids
- Rosanae
- Hamamelididae
- eurosids I
- Monochlamydeae
- Archichlamydeae
- Juglandanae

One of these is flagged "Preferred rank" (fabids) and the others are
"Normal rank". As a result only the rabies appear in the list of
parents.


These are simply notes to myself about taxonomic classifications in Wikidata. Classifications in Wikidata can be complex and are often not trees. For example, if we trace the parents of the frog family Leptodactylidae back we get a graph like this: Each oval represents a taxon in Wikidata, and each arrow connects a taxon to its parent(s) in Wikidata.

Messages de Rogue Scholar

Towards visualising classifications from Wikidata