A Comparison of Neural Models for Word Ordering

TL;DR

This paper compares various neural models for word ordering, introduces a new attention-based model, and demonstrates significant improvements in speed and quality on German and English datasets.

Contribution

It proposes a novel bag-to-sequence neural model for word ordering and a new search strategy, outperforming previous models in both accuracy and efficiency.

Findings

The new model significantly outperforms existing models on German data.

The proposed search strategy improves word ordering results.

The model achieves state-of-the-art performance on English Penn Treebank.

Abstract

We compare several language models for the word-ordering task and propose a new bag-to-sequence neural model based on attention-based sequence-to-sequence models. We evaluate the model on a large German WMT data set where it significantly outperforms existing models. We also describe a novel search strategy for LM-based word ordering and report results on the English Penn Treebank. Our best model setup outperforms prior work both in terms of speed and quality.