Semantic locality checker is broken

[tsarkov]

Semantic locality implementation is incorrect. What it should do it to take an axiom, replace all the entities not in signature with top/bottom entities of a kind and check whether the reasoner entails this new axiom. In FaCT++ I have to follow the root it is currently implemented because there is no isEntailed() interface there. Also this implementation misses the axiom transformation bit, which in FaCT++ is hidden in class TExpressionTranslator. I might implement this at some point but I'm not sure when I'll have time.

[sesuncedu]

The whole Atomic space Moduli-ser code could do with a bit of TLC. For example, there is the package named "decomposition", which is not java friendly. I assume that the Semantic Locality checker has been neglected since the 'cheap is good' stuff? Also, the performance of the atomic decomposer to an ontology with ~512K axioms is somewhat slow, even with syntactic locality (the result of assigning names to all complex class expressions in the extended Gene Ontology, reasoning over with ELK, and adding all inferred subclasses and quick disjoints). The quick fix is to not do that.

[ignazio1977]

The first thing it needs is proper unit tests - which unfortunately are an incredible pain to write. We'll have to bite the bullet and write them, at some point. Might be a good occasion to learn Cucumber or Fitnesse.