Lexical Translation Model Using a Deep Neural Network Architecture

TL;DR

This paper introduces a neural network-based lexical translation model that leverages global source context and shared parameters to improve translation quality, reducing data sparsity issues and achieving up to 0.5 BLEU point improvements.

Contribution

It presents a novel deep neural network architecture for lexical translation that integrates global context and shared parameters, enhancing translation performance over previous models.

Findings

Achieved up to 0.5 BLEU point improvement on TED translation tasks.

Effectively reduces data sparsity through shared parameters.

Leverages non-linear dependencies between source words.

Abstract

In this paper we combine the advantages of a model using global source sentence contexts, the Discriminative Word Lexicon, and neural networks. By using deep neural networks instead of the linear maximum entropy model in the Discriminative Word Lexicon models, we are able to leverage dependencies between different source words due to the non-linearity. Furthermore, the models for different target words can share parameters and therefore data sparsity problems are effectively reduced. By using this approach in a state-of-the-art translation system, we can improve the performance by up to 0.5 BLEU points for three different language pairs on the TED translation task.