---
id: python-crfsuite
version: "0.9.12"
license: MIT License
license_treatment: permissive
maintenance: aging
---

# python-crfsuite — Python binding for CRFsuite

License: permissive · Maintenance: aging · Downloads: 6.3M/mo

## What it is and what it does

python-crfsuite wraps the CRFsuite C/C++ library using Cython, providing a Python interface for conditional random field models. It is designed for sequence labeling tasks where you need to predict structured outputs based on features extracted from sequential data.

The package bundles CRFsuite and has no external runtime dependencies, making it lightweight to install. It supports Python 3.10 through 3.14 and is available as prebuilt wheels on major platforms. The binding exposes a trainer for fitting models and a tagger for inference, with a simpler codebase than some alternatives while remaining faster than the official SWIG wrapper.

Use it for:

- Named entity recognition: tag person, organization, location, and other entity types in text
- Part-of-speech tagging: label each word in a sentence with its grammatical role
- Information extraction: extract structured fields from unstructured text
- Chunking and segmentation: identify noun phrases, verb phrases, or other syntactic units
- Biomedical text mining: tag genes, proteins, diseases, or other domain entities in scientific literature

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

Python binding to CRFsuite for conditional random field sequence labeling and structured prediction.

Yes, if you need a lightweight, dependency-free conditional random field library for sequence labeling. The package is stable, permissively licensed, and well-supported on modern Python versions. Install friction is moderate due to compiled extensions, but prebuilt wheels mitigate this for common platforms. Maintenance is aging but the repository remains active; suitable for established production use.

## Install

pip install python-crfsuite
uv add python-crfsuite
poetry add python-crfsuite

## Installing python-crfsuite

Before you install: Medium install friction due to compiled extensions, but prebuilt wheels are available for Python 3.10–3.14 on macOS, Linux (manylinux and musl), and Windows. Last release was 234 days ago; repository is active and not archived, though maintenance status is aging.

License in practice: MIT-licensed package with a bundled BSD-licensed CRFsuite C/C++ library. Both are permissive, so commercial and proprietary use is permitted without restriction.

Quickstart:

pip install python-crfsuite

import python_crfsuite

trainer = python_crfsuite.Trainer(verbose=False)
trainer.append(xseq, yseq)
trainer.train('model.crfsuite')

Requires Python >=3.10. Prebuilt wheels exist for common platforms; source builds require a C/C++ compiler.

Verify before relying:

- Whether the package is actively maintained or in maintenance-only mode given the 'aging' status
- Performance characteristics compared to alternative CRF implementations or wrappers
- Specific API surface and available methods beyond trainer and tagger interfaces

## Package facts

- License: MIT License (permissive)
- Python support: supports_current
- Install friction: medium
- Maintenance: aging
- Downloads: 6.3M/month (top 5,000 on PyPI)
- Known vulnerabilities: none known

## Tags

conditional random fields python, sequence labeling library, named entity recognition, CRF structured prediction, python crfsuite binding, text tagging model, information extraction, sequence-labeling, structured-prediction, nlp

[View on SkillFed](https://skillfed.io/packages/python-crfsuite) · [View on PyPI](https://pypi.org/project/python-crfsuite/)