andrewyng / translation-agent

MIT License
4.83k stars 553 forks source link

Recommendation to Include "DUAL-REFLECT" paper in the Related Work Section #19

Open andongBlue opened 5 months ago

andongBlue commented 5 months ago

We would like to express our gratitude for this work, which has opened new avenues for using Large Language Models (LLMs) in the field of translation, bringing new vitality to the Machine Translation (MT) domain.

The translation approach mentioned in the README is worth delving into: prompting an LLM to translate a text from the source language to the target language, having the LLM reflect on the translation to provide constructive suggestions for improvement, and then using these suggestions to enhance the translation. This approach is consistent with the methodology in the paper “DUAL-REFLECT: Enhancing Large Language Models for Reflective Translation through Dual Learning Feedback Mechanisms” (https://arxiv.org/abs/2406.07232).

In this paper, the reflection process incorporates considerations of MT characteristics, utilizing the dual nature of MT to provide effective feedback for the iterative optimization process. This method offers a new perspective and pathway for improving translation quality.

As mentioned in the README, "A few academic research groups are also starting to look at LLM-based and agentic translation. We think it’s early days for this field!" Therefore, this work, being an early contribution to the field, holds significant reference value. It is highly recommended to include this work in the Related Work section of the project, as it can be beneficial to everyone.