Search in Constraint-Based Paraphrasing

Existing paraphrase systems either create successive drafts from an underlying representation , or they aim to correct texts. A diierent kind of paraphrase, also reeecting a real-world task, involves imposing, on an original text, external constraints in terms of length, read-ability, etc. This sort of paraphrase requires a new framework which allows production of a new text satisfying the constraints, but with minimal change from the original. In order for such a system to be feasible when used on a large scale, searching the space of candidate solution texts has to be made tractable. This paper examines a method for pruning the search space via branch-and-bound, evaluates three variants to nd the most eecient model, and discusses its relation to standard heuristic methods such as genetic algorithms.