Augmenting Large Language Model Translators via Translation Memories

TL;DR

This paper explores enhancing large language model translators by incorporating translation memories as prompts, significantly improving translation quality to rival state-of-the-art systems with less data.

Contribution

It demonstrates that prompting LLMs with high-quality translation memories enhances their translation performance, approaching the effectiveness of specialized neural machine translation models.

Findings

TM-based prompts improve LLM translation quality

Results are comparable to state-of-the-art NMT systems

Prompt understanding is crucial for leveraging TMs effectively

Abstract

Using translation memories (TMs) as prompts is a promising approach to in-context learning of machine translation models. In this work, we take a step towards prompting large language models (LLMs) with TMs and making them better translators. We find that the ability of LLMs to ``understand'' prompts is indeed helpful for making better use of TMs. Experiments show that the results of a pre-trained LLM translator can be greatly improved by using high-quality TM-based prompts. These results are even comparable to those of the state-of-the-art NMT systems which have access to large-scale in-domain bilingual data and are well tuned on the downstream tasks.