Precisiating natural language for a question answering system

We report on an application of Precisiated Natural Language (PNL) concepts and protoformal deduction, which are integral to Computational Theory of Perception, and Computing with Words, as developed by Lotfi Zadeh. A semi-automated precisiation process is part of an information extraction module for a question answering system. Simplified natural language statements (containing a single verb phrase) are first subjected to part-of-speech tagging to identify the verb phrase, subject phrase and the object phrase, if any. If the verb phrase is an "is-form" (covering all modalities and tenses of the "to be" verb) we dwell into further analysis of this sentence being one of the various PNL protoforms, such as X isr A, Y isr (X+B), QAs are Bs, and f(X) is A. Via protoformal deduction, more precise answers can be computed for a subset of a knowledge corpus (e.g. critical or frequently- asked topics) where fuzzy set definitions of vague terms are provided. For sentences without an "is-form" verb phrase, supplemental analyses detect causal facts, if-then rules, procedures, or simple propositions, and phrase- based deduction is subsequently applied where possible. Analyses are extended to query-type classification which is used to refine answer ratings.