Unifying Instance-level and Type-level QP Frames for Natural Language Understanding

1 One important role for qualitative representations is as a constituent of natural language semantics. The incremental nature of language means that information about situations and models typically arrives piece by piece, and must be assembled in order to create a qualitative model that can be reasoned with. Our prior research created distinct QP frame systems for instance-level and type-level qualitative models. This is inelegant as well as problematic, since the choice of constructing instance versus type level models ought to be made based on the text and task context, not a priori. This paper describes the design of a unified QP frame system, where most of the frame contents are agnostic with respect to whether the final description will be instance-level or type level. Clues from the NLU system’s analysis plus context will determine whether the final model constructed is instance level, type level, or a mixture of the two.