XCOMET

关于

This tool functions as an MCP (Model Context Protocol) Server, specifically designed to empower AI agents with advanced machine translation quality evaluation capabilities. It leverages Unbabel's xCOMET (eXplainable COMET) model to deliver comprehensive insights, including numerical quality scores, identification of error spans with severity levels, and efficient batch processing of multiple translation pairs. By integrating into AI agent workflows, it enables automated assessment and iterative improvement of translated content, ensuring higher linguistic quality in AI-driven applications.

主要功能

• Configurable xCOMET model (XL or XXL) for varying accuracy and resource needs
• Translation quality scoring (0-1)
• Seamless integration via Model Context Protocol (MCP)
• Detailed error detection with severity levels (minor/major/critical)
• 0 GitHub stars
• Efficient batch processing of translation pairs

使用案例

• Automated evaluation of machine translation outputs in AI agent workflows
• Integrating translation quality assessment into a multi-step translation pipeline
• Enabling AI agents to decide on accepting or retrying translations based on quality scores