A highly literate approach to Ontology Building
Phillip Lord and Jennifer Warrendar, Newcastle University
Highly Literate Ontology
What is an ontology and what is it for
The need for description
Ontologies and Programming
Literate Ontologies
Lenticular Text
The Future
What is an ontology
A specification of a conceptualisation
A conceptualisation of something
Lots of disagreement on what "something" is
What is the problem?
Clear that "something" is complicated
We all agree that it’s worth writing about.
Story
From OBI
A supernatant role is a role which inheres in a material entity…
— OBI
A pellet is a material entity which results from the aggregation of cells…
— OBI
Story
Pellet is a material entity?
Supernatant is a role?
Why the asymmetry?
Difficult to remember
Conclusions
Ontologies need a lot of description.
OWL doesn’t provide rich support for documentation
Annotations are un-ordered
Protege treats documentation as an add-on.
Literate Programming
Some programs require lots of documentation
"Literate Programming"
Single source file, generate out programmatic and documentation source
Best Examples: LaTeX and Sweave
Can we do Literate Ontology?
Ontologies need more documentation that most programs
Literate Ontology sounds like a good idea.
I’ve tried a variety of solutions.
All based around OWL Manchester Syntax
http://www.russet.org.uk/blog/1204
http://www.russet.org.uk/blog/1213
http://www.russet.org.uk/blog/1258
http://www.russet.org.uk/blog/1269
Conclusion
No one is going to write ontologies this way
Writing ontologies is hard
Only a crazy person would write an ontology in a flat-file
Conclusion
No one is going to write ontologies this way
Writing ontologies is hard
Only crazy person would write an ontology in a flat-file
Even GO has mostly stopped doing this now
But what about karyotypes?
A diversion: Karyotypes and Tawny-OWL
Humans (normally) have 46 chromosomes
Lost chromosomes or bits of chromosome are bad
Describing this ontology seems useful
A diversion: Karyotypes and Tawny-OWL
Each of the 23 pairs has bands (visible structures)
That’s 1000 classes all very similar
Protege is not going to work
Tawny-OWL
Interactive environment, create new ontologies, classes
Built on Clojure, the JVM, OWL API and some reasoners!
Tawny-OWL
Allows us to repurpose software engineering tools.
version control
commit discussions
pull management
unit testing
continuous integration
collaborative editing
Integrated Development Environment
Power Editor
Literate Ontology
We can write ontologies in text, using a rich programming environment
We can write documentation in text, also with a rich markup
Problem
Do we used a rich environment for programming?
Eclipse, Emacs CIDER, Intellij
Do we use a rich environment for documentation?
Scientific Word, LyX, Emacs auctex
Solution: Lenticular Text
Invented the notion of "Lenticular Text"
After "Lenticular Printing"
Text which changes depending on the way you look at it
Model-View-Controller for Text
Solution: Lenticular Text
View either representation
Edit either in a rich environment
Implementing Lenticular Text
Implemented this with Emacs "lentic" package
Why Emacs?
Strong support for Clojure
Strong support for several markup languages (latex, asciidoc, markdown).
Implementing Lenticular Text
Implementation is for Emacs
Algorithms are portable
Solution is very low-level
So, (mostly) neutral to text
Currently works for
Clojure, Emacs-lisp, Scala, Haskell, python
Latex, Org-Mode, Asciidoc
New languages can be added
What it looks like
Harder to explain than show
In practice, works straight-forwardly
What does it look like
Allows some fairly advanced editing
This is the pizza ontology
Both the ontology and documentation being written in Arabic
Right-to-Left text.
What does it look like
Here used over non-ontological source code
Using "Org-Mode", which is a markup language
Multi-lingual
With web publishing
What are we using it for
Karyotype Ontology
Tawny-OWL documentation (including tutorial yesterday)
Non-ontological uses (including lentic documentation)
Limitations
The tooling is complex (but powerful)
Different parts of the tooling have different limitations
LaTeX interacts poorly with web (but you can use asciidoc)
Asciidoc has weak referencing capabilities (but you can use latex)
Lentic is Emacs-specific
Summary
Both ontology and documentation should be considered first-class
This is possible with a text representation
Tawny-OWL makes ontology development in text possible
Many existing text formats for documentation
Lenticular Text allows editing both together
Future
New workflows for ontology development
Can we integrate office software?
Karyotype ontology has excel-driven test suite
What do we actually need to write?
Semantic documents as well as literate ontology
Acknowledgements
Jennifer Warrendar — The Karyotype Ontology
Aisha Blfgeh — Arabic Pizza
Feng Shu (Tumashu) — Chinese Screenshots
Phillip Lord — Tawny-OWL, Lentic
URLs
http://ukontology.org
UK Ontology Network 2015
http://github.com/phillord/tawny-owl
http://github.com/phillord/lentic
http://www.russet.org.uk/blog
