UnBIASing 1.0

A package to detect and debias text using pretrained models

UnBIASing

UnBIASing is a Python package that classifies, detects, and debiases textual content to promote unbiased information. By leveraging advanced machine learning models, UnBIASing provides users with tools to analyze and correct biases in their texts.

Features

• Bias Classification: Classifies textual content based on its bias using state-of-the-art models.

• Named Entity Recognition (NER): Detects named entities within the text that might be indicative of bias.

• Text Debiasing: Provides unbiased or debiased versions of the input text using an ensemble of advanced models.

Installation

pip install UnBIASing

Usage

Here's a basic example of how to use the BiasPipeline from UnBIASing:

from unbias import BiasPipeline

pipeline = BiasPipeline()
texts = ["Your sample text goes here."]
classification_results, ner_results, debiaser_results = pipeline.process(texts)

# If you wish to print the results
pipeline.pretty_print(texts, classification_results, ner_results, debiaser_results)

# Convert results to a Pandas DataFrame
df = results_to_dataframe(texts, classification_results, ner_results, debiaser_results)
print(df)

Dependencies

• Transformers
• Torch
• Pandas
• SentencePiece

License

MIT

---

We hope UnBIASing proves useful in your journey to make the digital world a more inclusive and unbiased space. For any queries or feedback, feel free to Shaina Raza at [e-mail]

---