Dependency Graph-to-String Statistical Machine Translation

TL;DR

This paper introduces graph-based translation models that convert dependency graphs into target language strings, improving translation quality over sequence and tree-based methods, especially in handling non-syntactic phrases and reordering.

Contribution

The paper presents novel graph-to-string translation models, including a segment-based approach and a recursive grammar model, with efficient parsing implementations and demonstrated improvements in Chinese-English and German-English translation.

Findings

Graph-based models outperform sequence and tree baselines.

Recursive grammar model improves phrase reordering.

Models are effective on Chinese-English and German-English tasks.

Abstract

We present graph-based translation models which translate source graphs into target strings. Source graphs are constructed from dependency trees with extra links so that non-syntactic phrases are connected. Inspired by phrase-based models, we first introduce a translation model which segments a graph into a sequence of disjoint subgraphs and generates a translation by combining subgraph translations left-to-right using beam search. However, similar to phrase-based models, this model is weak at phrase reordering. Therefore, we further introduce a model based on a synchronous node replacement grammar which learns recursive translation rules. We provide two implementations of the model with different restrictions so that source graphs can be parsed efficiently. Experiments on Chinese--English and German--English show that our graph-based models are significantly better than corresponding sequence- and tree-based baselines.