Briscoe et al. [17] present a formal account of how to model defaults in the lexicon while still allowing the defaults to be superceded. At issue is how to allow for blocking of lexical rules (for inheritance networks) in certain situations, such as when another lexical item is equivalent. For instance, animal grinding is the rule of extending the name for an animal to also cover the food for an animal, as in ``lamb''. This is normally blocked for ``cow'' since another word, ``beef'', already accounts for it. Previous approaches accounting for this have tended to be too inflexible, as in always blocking in certain situations, or requiring too much redundancy in the specifications to allow for the exceptions. The solution uses default logic, specifically conditional entailment, to model nonmonotonic reasoning. One advantage of this logic over most other default logics is its support for conflict resolution.
Kayser and Abir [64] also present a nonmonotonic approach for modeling sense extensions. They base this approach on Reiter's default logic. This is illustrated with the handling of the extension of the name for an action to apply to the object as well. Blocking is addressed by explicitly adding clauses that disable the default rule inferencing.
Rule conflict is handled by assigning qualitative labels to the results and using an application-specific selection mechanism. In contrast, Briscoe et al. [17] explicitly model rule priorities through specificity constraints.