On LM Heuristics for the Cube Growing Algorithm

Current approaches to statistical machine translation try to incorporate more structure into the translation process by including explicit syntactic information in form of a formal grammar (with a possible, but not necessary, correspondence to a linguistic motivated grammar). These more structured models incur into an increased generation cost, and efficient algorithms must be developed. In this paper we concentrate on the cube growing algorithm, a lazy version of the cube grow algorithm. The efficiency of this algorithm depends on a heuristic for language model computation, which is only scarcely discussed in the original paper. In this paper we investigate the effect of this heuristic on translation performance and efficiency and propose a new heuristic which efficiently decreases memory requirements and computation time, while maintaining translation performance.