OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Evaluation & Observability

Comet — Unbabel

COMET is an open-source Python framework that evaluates machine-translation quality with neural metrics, including reference-based, reference-free, contextual and explainable evaluation models.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Comet — Unbabel — OpenViST project profile in AI Evaluation & Observability.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use COMET after producing machine-translation candidates when quality scoring, system comparison, error analysis or candidate selection is required.

Who it is for

  • Machine-translation researchers
  • Developers comparing translation systems
  • Teams analyzing translation errors
  • Researchers evaluating document or chat translation quality

Documented uses

  • Score machine-translation outputs against references or source text
  • Compare several translation systems with statistical significance testing
  • Identify and categorize translation errors with XCOMET
  • Evaluate translations using document context
  • Rank sampled translations with minimum Bayes risk decoding
Before you choose

Requirements & limits

COMET is presented as a public Apache-2.0-licensed repository. Python 3.8 or newer is required; some models require Hugging Face Hub license acknowledgement and login.

  • Choose reference-based models when trusted reference translations are available.
  • Use reference-free models when evaluation must rely on source text and the candidate translation.
  • Choose XCOMET when categorized error findings and JSON export are useful.
  • Use context-enabled evaluation when discourse or document context affects translation quality.
  • Consider the required local Python environment and model-specific Hugging Face access conditions.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Install COMET for local evaluation

Set up the COMET Python package or a local development checkout.

Before you start

  • Python 3.8 or newer.
  • For local development, Git and Poetry.
  • Some models may require Hugging Face Hub license acknowledgement and login.

Steps

  1. Install or upgrade pip and install the unbabel-comet package from PyPI.

  2. For local development, clone the COMET repository, enter its directory, install Poetry and run the documented Poetry installation.

  3. If using a model with a Hugging Face Hub licensing requirement, acknowledge the model license and log in to Hugging Face Hub.

Expected outcome

A COMET environment is prepared for the documented command-line or development workflows.

Guide limitations

  • This setup recipe does not verify a successful installation or model execution.
  • Model-specific access requirements vary.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
4 October 2026
Original website title
Unbabel/COMET · Official homepage title · retained in its source language
How we review sourcesSuggest a correction