Efficient Terminology Integration for LLM-based Translation in Specialized Domains

TL;DR

This paper presents a new method for training large language models to better handle specialized terminology in translation tasks, using efficient term extraction and data reconstruction techniques, resulting in high-quality, consistent translations in domain-specific fields.

Contribution

The paper introduces a systematic approach combining Trie Tree-based term extraction and data reconstruction to improve terminology integration in LLM-based translation with less data.

Findings

Achieved highest translation score in WMT patent task

Enhanced model's ability to handle specialized terminology

Demonstrated broad applicability in domain-specific translation

Abstract

Traditional machine translation methods typically involve training models directly on large parallel corpora, with limited emphasis on specialized terminology. However, In specialized fields such as patent, finance, or biomedical domains, terminology is crucial for translation, with many terms that needs to be translated following agreed-upon conventions. In this paper we introduce a methodology that efficiently trains models with a smaller amount of data while preserving the accuracy of terminology translation. We achieve this through a systematic process of term extraction and glossary creation using the Trie Tree algorithm, followed by data reconstruction to teach the LLM how to integrate these specialized terms. This methodology enhances the model's ability to handle specialized terminology and ensures high-quality translations, particularly in fields where term consistency is crucial. Our approach has demonstrated exceptional performance, achieving the highest translation score among participants in the WMT patent task to date, showcasing its effectiveness and broad applicability in specialized translation domains where general methods often fall short.