Memorizing Read Ontologies

Memorization allows tawny to cope with change in ontology dependencies, by tracking the identifiers that tawny has used in the past, so that it can recognise changes in the future. This is important for reading ontologies which use hard-to-remember identifiers.

The problem

The problem that memorize addresses was first described in an article. Many ontologies use numeric identifiers; this are hard to use computationally, because they are hard for the developer to remember. Tawny can provide an arbitrary transformation of OWL annotation properties to produce a readable label. However, downstream ontologies may change these labels which will break any ontologies build using these.

In a GUI environment, this causes less problems because there is a separation of the underlying model and what the user sees; if the label changes, and the identifier does not, a GUI will just show the user the update. This is less easy to achieve with a code base, but tawny manages this with memorise.clj.

Solution

The memorization functionality stores the mappings between label and identifiers when ever a downstream OWL file is read. When a file is read for a second time, Tawny checks these mappings to see if any of the identifiers has a new label. In this case, tawny effectively generates an alias to the entity using both the old and new label. However, use of the old label will result in a deprecation warning, printed to the repl. This means that downstream ontologies will still work, while allowing developers to migrate to new identifiers.

In Practice

Use of memorise is relatively straight-forward. The upstream ontology needs to be read as described previously. To save the mappings that have been made, use the memorize function:

(tawny.memorise/memorise obi "./src/tawny/obi_memorise.clj")

The location for the "memorise" is variable, but I use the same directory as the load file because it will be versioned along with other source. If you are using a downloaded version of the OWL source, this function only needs to be run every time it is updated.

To remember the mappings made previously, we use the counter-part remember function.

(tawny.memorise/remember obi "./src/tawny/obi/obi_memorise.clj")

Memorise generates a print statement for all the obsolete mappings that it finds, and will generate another print statement whenever the obsolete term is actually used.

Next

• OBO style, numeric identifiers