Improved statistical machine translation using monolingual paraphrases

TL;DR

This paper introduces a monolingual paraphrasing technique that enhances statistical machine translation by generating additional training data from existing sentences, leading to significant performance improvements.

Contribution

It presents a novel recursive paraphrasing method based on syntactic trees to augment training data without additional data collection.

Findings

Achieved 33-50% of the gains from doubling training data

Improved translation quality using paraphrased monolingual data

Demonstrated effectiveness of syntactic paraphrasing in SMT

Abstract

We propose a novel monolingual sentence paraphrasing method for augmenting the training data for statistical machine translation systems "for free" -- by creating it from data that is already available rather than having to create more aligned data. Starting with a syntactic tree, we recursively generate new sentence variants where noun compounds are paraphrased using suitable prepositions, and vice-versa -- preposition-containing noun phrases are turned into noun compounds. The evaluation shows an improvement equivalent to 33%-50% of that of doubling the amount of training data.