Babel user information
ja-N この利用者は日本語母語としています。
en-1 This user has basic knowledge of English.
QuickStatements logoThis user uses QuickStatements.
ȹThis user uses quickpresets, with custom settings.
Wikidata cube almost solved.svgThis user uses Mix'n'match.
GeneWikidata-logo-en.pngThis user is a member of WikiProject Molecular biology.
Blutkreislauf.pngThis user is a member of
WikiProject Anatomy
Wikidata Medicine Task Force Userbox.svgThis user is a member of
WikiProject Medicine
Users by language

What kind of "relation" is used to construct tree in various ontologiesEdit

Relations (black arrows) connect items (blue dots) in certain manner.

In ontology, "relation" defines relationship of 2 items. Different items are connected by relation. Visually thinking, relation is corresponding to edge in graph (black arrows in the diagram at right). In Wikidata, for example, this data (Q42#P3373) connects different person in certain manner.

By collecting these relations, mesh (or tree) is constructed.


You can see their tree at here[1]. TA doesn't mention what kind of relation they indicate. They simply mentioning as "relationship".

Indenting and styles of heading are used to indicate the relationship of one term to another.
—TA documentation: TA User Guide Last revision: 5 october 2010 (pdf)

Cornelius Rosse (2001) state that TA hierarchical structure indicates "various relationships." He said the relationship is sometimes "is a", and sometimes "has part of".

A hierarchy of headings is presented through the use of different fonts and varied shading of the panels in which the headings appear. Combined with indentations in the list, these headings have the potential for representing various relationships between the terms.
—Rosse C (2001) Terminologia anatomica: considered from the perspective of next-generation knowledge sources. (pdf)


FMA would be the most sophisticated human anatomy ontology. They uses various properties including relationships. Their philosophy is explained briefly at here[2]. As of Oct. 2017, they uses 168 properties[3].You can see whole list at here[4]. Among them, they draw tree based on <superclass ↔ subclass> relations.

In official browser, following 4 relations are used as basic component of the interface.

In Mejino JLV (2003), they explains about the regional/constitutional part relations as follows.

Constitutional part is a primary partition of an anatomical structure into its compositionally distinct anatomical elements. In the context of the whole, an element is any relatively simple component of which a larger, more complex anatomical structure is compounded; i.e., the partition is compositional rather than spatial.

Regional part is a primary partition that spatially subdivides an anatomical structure into sets of diverse constitutional parts that share a given location within the whole; i.e., the partition is spatial rather than compositional. Entities in all subclasses of Anatomical structure have both constitutional and regional parts
—Mejino JLV, et. al. Representing Complexity in Part-Whole Relationships within the Foundational Model of Anatomy. AMIA Annual Symposium Proceedings. 2003;2003:450-454. (PMC)


UBERON uses various properties including relationships. Its philosophy is explained briefly at here[5]. As of Oct. 2017, they uses 220 properties[6].You can see whole list at here[7]. Among them, it seems that UBERON drawing tree based on <part ofhas part> relations.

MeSH treeEdit

MeSH makes hierarchical tree basically for convenience for users. They don't define in strict sense what kind of hierarchy it is. There are brief explanation about MeSH tree at here[8] or here[9]. Following is excerpt.

These trees should not be regarded as representing an authoritative subject classification system but rather as arrangements of descriptors for the guidance and convenience of persons who are assigning subject headings to documents or are searching for literature.


Each MeSH descriptor appears in at least one place in the trees, and may appear in as many additional places as may be appropriate.

MeSH Tree Structures Last updated: 06 December 2016


NeuroNames draw single hierarchical tree based on <superstructure ↔ substructure> relation, like this[10], this[11] or this[12]. Brief explanation about the tree is here[13] and here[14]


NeuroLex creates tree through "is a" relationship. Following is excerpt from FAQ.

NeuroLex was designed as a single tree hierarchy, that is, a given category has one and only one parent category (super category) and one or more children categories. A category is related to its parent and children through a "is a type of" relationship, also called an "is a" relationship. So, a Neuron "is a type of" Cell; a "Purkinje cell" is a type of" Neuron.

On Categories (or Concepts)Edit

Basic features of categories (or concepts)Edit

Good summary

  1. Categories are fuzzy
    1. In contemporary history of philosophy, most famous discussion about the fuzziness of concept might be w:family resemblance by w:Ludwig Wittgenstein.
    2. It was developed into w:Prototype theory in psychology.
  2. Categories are learned
  3. Categories are flexible, and goal-oriented
  4. Categories and category terms are interwoven

Where is categories (or concepts)?Edit

Very very roughly speaking, in current scientific theory, it is thought that concepts are located at ATL (anterior temporal lobe), or somewhere around there.

This view is called "semantic hub model". See wikiversity:Psycholinguistics/Semantics in the Brain.

Implication to our project (tree or mesh)Edit

What do these things implicate to our project Wikidata?

These nature imposing us one dilemma.

We have two distinguishable desires which toward different directions.

  1. First desire: "We want make beautiful tree data structure in Wikidata"
  2. Second desire: "We want to collect all possible knowledge"

If we pursuit first desire, we aim tree structure like this.

Beautiful w:tree structure.

But if we pursuit second desire, what happens? "All possible knowledge" must, in some degree, reflect our brain structure. In other words, I suppose that it would result in some kind of mesh-like topology, like w:connectome.

Beautiful mesh. Human w:connectome. Neural connections in the brain.

If you are thinking about only first one (tree structure), second point seems to you as "messy". Conversely, If you are thinking about only second one (mesh structure), first point seems to you "constrained" or "tight".

As far as I think, this conflict is basically unsolvable through one‐sided approach (to say, taking only one side). In my current thinking, best approach is "balancing these two desire in pragmatic point of view".

Deeper implicationEdit

If you have philosophical trait of thinking, you would notice that there are more deeper implication from this dilemma. It is thoughts related to w:Platonism.

If you believe following two things:

  1. There is true structure of nature.
  2. It is descriptivable in short sized-data.

I suppose you are automatically to be attracted to the following thought.

  • Wikidata can and should reflect the true structure of nature.

I could call this desire as Platonic wish.

To be continued...

Related Wikidata pagesEdit

Tree testEdit

By using Template:Tree, we get the following tree automatically.


Starting from human chromosome 1 (Q430258), tracking through followed by (P156)


Starting from Alan Turing (Q7251), tracking through doctoral advisor (P184)


Starting from Alphabet Inc. (Q20800404), tracking through subsidiary (P355)


Starting from John F. Kennedy (Q9696), tracking through killed by (P157)


Starting from Amazon (Q3783), tracking through tributary (P974)