Introducing the NewsPaLM MBR and QE Dataset: LLM-Generated High-Quality Parallel Data Outperforms Traditional Web-Crawled Data

TL;DR

This paper introduces a high-quality, LLM-generated parallel dataset that surpasses traditional web-crawled data in training neural machine translation models, demonstrating significant performance improvements.

Contribution

The authors release the first LLM-generated, MBR-decoded, QE-reranked dataset with sentence and multi-sentence examples, showing its effectiveness in NMT training.

Findings

Training on the dataset outperforms larger web-crawled data.

Self-distillation of the LLM further improves translation quality.

High-quality machine-generated data enhances NMT performance.

Abstract

Recent research in neural machine translation (NMT) has shown that training on high-quality machine-generated data can outperform training on human-generated data. This work accompanies the first-ever release of a LLM-generated, MBR-decoded and QE-reranked dataset with both sentence-level and multi-sentence examples. We perform extensive experiments to demonstrate the quality of our dataset in terms of its downstream impact on NMT model performance. We find that training from scratch on our (machine-generated) dataset outperforms training on the (web-crawled) WMT'23 training dataset (which is 300 times larger), and also outperforms training on the top-quality subset of the WMT'23 training dataset. We also find that performing self-distillation by finetuning the LLM which generated this dataset outperforms the LLM's strong few-shot baseline. These findings corroborate the quality of our dataset, and demonstrate the value of high-quality machine-generated data in improving performance of NMT models.