Translation Agent

Preview：

Introduce：

Translation Agent is a machine translation demonstration project using reflective workflows. Developed by Andrew Ng and collaborators, it focuses on text translation using large language models (LLMS) and reflects on the translation results to suggest improvements. The model is highly customizable, enabling adaptation of translation styles, handling idioms and proper nouns, and optimization for specific regions or dialects. Although the software is not mature enough at present, it has shown the potential to be more competitive with traditional machine translation systems at times.

Stakeholders:
The target audience is developers, translators, and any individual or organization that needs machine translation services. Due to its high degree of customizability, the product is particularly suitable for users who need a specific translation style or deal with specialized terms.
Usage Scenario Examples:

• Developers use Translation Agent to translate technical documents from English to Spanish while maintaining consistency in terminology.
• The model is used by translation companies to customize translation services for multilingual clients and improve translation quality and efficiency.
• Academic researchers have used the model to study translation strategies and effects between different languages.

The features of the tool:

• Text translation from source language to target language using a large language model
• Reflect on the translation results and put forward some suggestions for improvement
• Optimize translation results according to suggestions
• Highly customizable, easy to adjust translation styles and deal with specific terms
• Supports translation optimization for specific regions or dialects
• Change the prompts to suit different translation needs

Steps for Use:

• Install the Poetry package manager, which is a prerequisite for running the workflow.
• Create an.env file containing OPENAI_API_KEY, refer to the.env.samlile file.
• Install dependencies with Poetry and activate the virtual environment.
• Import the translation_agent module and translate using the translate function, specifying the source language, target language, and country code.
• Look at examliles/examlile_scrilit.liy for a sample script to see how to use this model.

Tool’s Tabs: Machine translation, large language models