A recursive statistical translation model

  • Authors:
  • Juan Miguel Vilar;Enrique Vidal

  • Affiliations:
  • Universitat Jaume I, Castellón, Spain;Universidad Politécnica de Valencia, Valencia, Spain

  • Venue:
  • ParaText '05 Proceedings of the ACL Workshop on Building and Using Parallel Texts
  • Year:
  • 2005

Quantified Score

Hi-index 0.00

Visualization

Abstract

A new model for statistical translation is presented. A novel feature of this model is that the alignments it produces are hierarchically arranged. The generative process begins by splitting the input sentence in two parts. Each of the parts is translated by a recursive application of the model and the resulting translation are then concatenated. If the sentence is small enough, a simpler model (in our case IBM's model 1) is applied. The training of the model is explained. Finally, the model is evaluated using the corpora from a large vocabulary shared task.