Parsing with Commitment

Abstract

One of the most attractive features of logic programming is the possibility of executing logic programs in parallel. Recall that a logic program consists of a finite set of rules of the form:

$$A: - {B_1}, \ldots ,{B_n} n \geqslant 0$$

where A and the B_i are atoms (see chapter 5). Given a query such as:

$$? - Q$$

the sequential method of solution of the query is to find some rule in the program whose head atom unifies with Q, say

$$Q': - {R_1}, \ldots ,{R_k}$$

and replace the goal Q with the conjuction of goals R’₁,…,R’ _k where the primes denote that the substitution derived from the unification of Q and Q’ has been applied. If Q’ had been a unit rule, then the query would have been solved. If not, the process continues by selecting some R’ _j, finding a rule whose head unifies with that, replacing R’ _j with the body of the selected rule under the derived substitution, and so on, until either all subgoals have been solved, or until no further progress can be made. The Prolog programming language (as opposed to pure logic programming) always selects the first goal, proceeds from left to right in seeking a solution, and backtracks when it runs into an impasse. Prolog considers the rules which unify with a goal atom in strictly textual order.