As a user, I want a method for computing the COMET metric for a given translated text, its source text, and a reference translation; so that I can score machine translations.

[laurejt]

COMET is an LLM-based MT metric.

• Use HuggingFace's evaluate implementation of the metric

Inputs:

• tr_text: The text of the translation to be scored
• src_text: The original source text
• ref_text: The reference translation

Output: COMET score (float), ranges between 0 and 1.