Identifying Non-Redundant Literals in Clauses with Uniqueness Propagation

Several authors have proposed increasingly efficient methods for computing the least general generalization of two clauses under theta-subsumption. The most expensive step in this computation is the reduction of the clause to its minimal size. In general, this requires testing all literals for redundancy, which in the worst case requires a thetasubsumption test for each of them. In this paper, we present a small result that is complementary to earlier published results. It basically extends the conditions under which a literal can be decided non-redundant without running a theta-subsumption test. Preliminary experiments show a speedup factor of about 3 on one dataset when this new result is used.