Inconsistency Detection


Source Distribution

Detecting Inconsistencies in Large First-Order Knowledge Bases describes an approach to systematically extract many coherent subsets of axioms (“probes”) from a large knowledge base and test them for inconsistency. This succeeds much more readily than trying to uncover an inconsistency in the full axiomatisation.

  1. The probe generator has been integrated into e_axfilter, which is part of the standard E distribution. To make it available to interested parties before the next release of E, the version used in the paper can be downloaded here. Follow the normal installation instruction in the README file to build the system.

  2. E 2.0pre12 source distribution (~2.1 MB, .tgz)

  3. *** PLEASE NOTE: The updated tool is now part of the the E 2.0 release. This pre-release should only be used for exact reproduction of the experiments in the paper. ***

  4. A possible command line for experiments would be

  5. e_axfilter --seed-method=lda --seed-subsample=f200 --seed-symbols=p bigax.p

  6. where:

  7. --seed-method=lda selects the the Largest, the most Diverse, and All axioms for a given seed symbol as seed axioms

  8. --seed-subsample=f200 selects the 200 most frequent symbols as seed symbols

  9. -seed-symbols=p restricts the pool of candidate symbols to predicate symbols

  10. bigax.p is the name of the big axiomatisation to be sampled