An efficient generation algorithm for lexicalist MT

Abstract

The lexicalist approach to Machine Translation offers significant advantages in the development of linguistic descriptions. However, the Shake-and-Bake generation algorithm of (Whitelock, 1992) is NPcomplete. We present a polynomial time algorithm for lexicalist MT generation provided that sufficient information can be transferred to ensure more determinism.

Minimal Recursion Semantics: An Introduction

“…This avoids problems of the sort we have been considering, which essentially arise from the structure of the LF. However, from our perspective, this is not a good solution, because it is quite specific to MT and it imposes stringent conditions on the grammar (at least in the efficient form of Shake-and-Bake processing described by Poznanski et al, 1995). An alternative is to modify the form of the semantic representation, in particular to use a non-recursive, or flat representation such as those developed by Phillips (1993) or Trujillo (1995) (the latter uses flat semantic structures in conjunction with Shake-and-Bake processing).…”

A chart generator for Shake and Bake machine translation

“…However, our algorithm is preferable in cases where translation involves the creation of several possible target language bags, each of which does not necessarily result in the successful generation of a sentence. If one is able to assume that there is only one solution generated from a bag then there is a polynomial time algorithm [12], but enough information must be transferred from the source bag to the target sign to ensure that combination is deterministicthis requires additional restrictions in the grammars and lexicons. Our algorithm appears to be similar to the chart generator developed independently by Trujillo and very briefly described in [13].…”

“…For an inactive edge of the form <N, C [ω] → α. >, we can use a term of the form inactive_edge(N, C, Omega, Alpha) (12) while for an active edge of the form <M, A[ω] → α. C β>, we could use the term active_edge(M, A, Omega, Alpha, [C|Beta]).…”

Connectivity in bag generation

“…Since the computational complexity of the greedy bag generator (Poznański et al, 1995) is polynomial (i.e. O(n 4 )), the effect of redundant substructures is not as detrimental as for parser based generators.…”

“…

This paper presents a pruning technique which can be used to reduce the number of paths searched in rule-based bag generators of the type proposed by (Poznański et al, 1995) and (Popowich, 1995). Pruning the search space in these generators is important given the computational cost of bag generation.

…”