Skip to content

What is unbias-plus?

unbias-plus is an AI-driven toolkit for bias detection and debiasing in text. It identifies biased segments, classifies severity, suggests neutral replacements with reasoning, and produces a full neutral rewrite—supporting risk identification, mitigation, and more trustworthy text systems.

PyPI version Open In Colab GitHub — Setup, usage, CLI, and API.

Highlights

  • Bias detection — Pinpoints biased phrases in text and returns them as structured segments with character-level offsets for highlighting.
  • Classification — Overall binary label (biased/unbiased), per-segment severity (low/medium/high), and bias type (e.g. loaded language, framing).
  • Reasoning — Each segment includes an explanation of why it is considered biased, so you can review and act with context.
  • Debiasing — Per-segment neutral replacements plus a full rewritten unbiased_text of the input.
  • Structured output — Pydantic-validated BiasResult: binary_label, severity (1–5), biased_segments, and unbiased_text. Use via CLI (unbias-plus --text "..."), REST API (FastAPI + demo UI), or Python API (UnBiasPlus().analyze()).
  • Production-ready — Optional 4-bit quantization and default fine-tuned model vector-institute/Unbias-plus-Qwen2.5.

Video tutorials

Videos show how to use the demo UI when you run the API with unbias-plus --serve:

Quick start

uv sync
source .venv/bin/activate   # or .venv\Scripts\activate on Windows
unbias-plus --text "Women are too emotional to lead."

Or in Python:

from unbias_plus import UnBiasPlus
pipe = UnBiasPlus()  # uses default model, or pass model_name_or_path=
result = pipe.analyze("Women are too emotional to lead.")
print(result.binary_label, result.unbiased_text)

Contents

  • User Guide — Installation, CLI, REST API, Python API, and development.
  • API Reference — Module and class reference for unbias_plus.
  • Team — Contributors, acknowledgement, and support.
  • License — Apache License 2.0.