Computing with Words (CW or CWW) is a system of computation which offers an important capability that traditional systems of computation do not have—a capability to compute with information described in a natural language. In the main, CW is concerned with solution of problems which are stated in a natural language. The importance of CW derives from the fact that much of human knowledge is perception-based and is described in a natural language.
The point of departure in CW is a question, q, of the form: What is the value of a variable, X? q is associated with a question-relevant information set, I, an association expressed as X is I, meaning that the answer to q, Ans(q/I), is to be deduced (computed) from I. Typically, I consists of a collection of propositions, p1, …, pn, which individually or collectively are carriers of information about the value of X. In I, some or all of the pi, i=1, …, n, are expressed in a natural language. Some of the pi may be drawn from external sources of information, typically from world knowledge. I is open if it includes propositions drawn from external sources of information. I is closed if inclusion is not allowed.
Precisiation of meaning is a prerequisite to computation with information which is described in a natural language. If p is a proposition drawn from a natural language, then precisiation of p leads to a computation-ready proposition, p*, which may be viewed as a computational meaning of p or, equivalently, as a model of p. p* is assumed to be mathematically well-defined and is intended to serve as an object of computation. In CW, there are two levels of generality, Level 1 and Level 2. In Level 1, the carriers of information are words. In the more recent Level 2 CW, the carriers of information are words and propositions. Today, the bulk of the literature is still focused on Level 1 CW. In particular, the widely used calculi of fuzzy if-then rules fall within the province of Level 1 CW.
Computation of Ans(q/I) is carried out in two phases. Phase 1, called Precisiation, involves precisiation of q and I, leading to precisiated q, q*, and precisiated I, I*. Phase 2, called Computation, involves computation with q* and I*, leading to Ans(q/I). This is done through the use of an aggregation function which has the pi* as its arguments. In CW, the pi* are represented as generalized assignment statements or, equivalently, as generalized constraints. Computation with the pi* involves propagation and counterpropagation of generalized constraints. In CW, precisiation and computation employ the machinery of fuzzy logic.
CW has important applications to decision analysis, question-answering systems, system modeling, specification and optimization, and mechanization of natural language understanding. Basically, CW opens the door to a wide-ranging enlargement of the role of natural languages in scientific theories.