Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

General2Specialized LLMs Translation for E-commerce

Kaidi Chen, Ben Chen, Dehong Gao, Huangyu Dai, Wen Jiang, Wei Ning, Shanqing Yu, Libin Yang, Xiaoyan Cai

TL;DR

The paper tackles the poor performance of general NMT/LLMs on e-commerce translation by addressing domain-specific vocabularies and writing styles. It introduces G2ST, a two-step fine-tuning framework augmented by self-contrastive semantic enhancement, and constructs two domain resources: a 20k Chinese–English term-pairs vocabulary and a 7k Chinese product-title parallel corpus. By expanding vocabulary and guiding the model with domain-specific data, G2ST adapts general models (e.g., Qwen-14B) to e-commerce translation, achieving state-of-the-art results on real Alibaba titles and outperforming models like LLaMA, GPT-3.5, and even GPT-4 on several metrics. The approach demonstrates the value of domain resources and contrastive learning for robust, domain-specific translation and suggests applicability to multilingual MT in the future.

Abstract

Existing Neural Machine Translation (NMT) models mainly handle translation in the general domain, while overlooking domains with special writing formulas, such as e-commerce and legal documents. Taking e-commerce as an example, the texts usually include amounts of domain-related words and have more grammar problems, which leads to inferior performances of current NMT methods. To address these problems, we collect two domain-related resources, including a set of term pairs (aligned Chinese-English bilingual terms) and a parallel corpus annotated for the e-commerce domain. Furthermore, we propose a two-step fine-tuning paradigm (named G2ST) with self-contrastive semantic enhancement to transfer one general NMT model to the specialized NMT model for e-commerce. The paradigm can be used for the NMT models based on Large language models (LLMs). Extensive evaluations on real e-commerce titles demonstrate the superior translation quality and robustness of our G2ST approach, as compared with state-of-the-art NMT models such as LLaMA, Qwen, GPT-3.5, and even GPT-4.

General2Specialized LLMs Translation for E-commerce

TL;DR

The paper tackles the poor performance of general NMT/LLMs on e-commerce translation by addressing domain-specific vocabularies and writing styles. It introduces G2ST, a two-step fine-tuning framework augmented by self-contrastive semantic enhancement, and constructs two domain resources: a 20k Chinese–English term-pairs vocabulary and a 7k Chinese product-title parallel corpus. By expanding vocabulary and guiding the model with domain-specific data, G2ST adapts general models (e.g., Qwen-14B) to e-commerce translation, achieving state-of-the-art results on real Alibaba titles and outperforming models like LLaMA, GPT-3.5, and even GPT-4 on several metrics. The approach demonstrates the value of domain resources and contrastive learning for robust, domain-specific translation and suggests applicability to multilingual MT in the future.

Abstract

Existing Neural Machine Translation (NMT) models mainly handle translation in the general domain, while overlooking domains with special writing formulas, such as e-commerce and legal documents. Taking e-commerce as an example, the texts usually include amounts of domain-related words and have more grammar problems, which leads to inferior performances of current NMT methods. To address these problems, we collect two domain-related resources, including a set of term pairs (aligned Chinese-English bilingual terms) and a parallel corpus annotated for the e-commerce domain. Furthermore, we propose a two-step fine-tuning paradigm (named G2ST) with self-contrastive semantic enhancement to transfer one general NMT model to the specialized NMT model for e-commerce. The paradigm can be used for the NMT models based on Large language models (LLMs). Extensive evaluations on real e-commerce titles demonstrate the superior translation quality and robustness of our G2ST approach, as compared with state-of-the-art NMT models such as LLaMA, Qwen, GPT-3.5, and even GPT-4.
Paper Structure (11 sections, 4 equations, 1 figure, 4 tables)

This paper contains 11 sections, 4 equations, 1 figure, 4 tables.

Table of Contents

  1. Introduction
  2. Method
  3. Domain-related Resources
  4. General to Special Transfer
  5. Self-contrastive Semantic Enhancement
  6. Experiments
  7. Experimental Settings
  8. Main Results
  9. Ablation Study
  10. Related Work
  11. Conclusion

Figures (1)

  • Figure 1: An overview of our G2ST approach. TP = Terms Pairs. PC = Parallel Corpus. The blue box represents the Self-contrastive module, the gray circle indicates that this node is dropped.