AI-Generated Text Verifier

What It Does

Installation

pip install transformers==4.33.3 scikit-learn==1.3.0 numpy==1.26.0

Usage

from ai_text_verifier import verify_text

text = "Generated by AI."
score = verify_text(text)
print(f"Probability of being AI-generated: {score:.2f}")

Source Code

import numpy as np
from transformers import pipeline
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.linear_model import LogisticRegression

def verify_text(text: str) -> float:
    """
    Analyzes the given text and returns a probability score indicating
    whether the text is AI-generated.

    Args:
        text (str): The text to analyze.

    Returns:
        float: A probability score (0.0 to 1.0) where higher values indicate
               the text is more likely AI-generated.
    """
    if not text or not isinstance(text, str):
        raise ValueError("Input must be a non-empty string.")

    # Load a transformer-based model for text embeddings
    try:
        embedding_pipeline = pipeline("feature-extraction", model="distilbert-base-uncased")
    except Exception as e:
        raise RuntimeError("Failed to load transformer model.") from e

    # Generate embeddings for the input text
    try:
        embeddings = embedding_pipeline(text)
        embeddings = np.mean(embeddings[0], axis=0)  # Average pooling
    except Exception as e:
        raise RuntimeError("Failed to generate embeddings.") from e

    # Simulate a trained model (Logistic Regression for simplicity)
    # Normally, you'd load a pre-trained model here
    vectorizer = TfidfVectorizer()
    classifier = LogisticRegression()

    # Mock training data for demonstration purposes
    mock_texts = ["This is a human-written sentence.", "Generated by AI."]
    mock_labels = [0, 1]  # 0 = human, 1 = AI

    try:
        tfidf_features = vectorizer.fit_transform(mock_texts)
        classifier.fit(tfidf_features, mock_labels)
    except Exception as e:
        raise RuntimeError("Failed to train mock classifier.") from e

    # Predict authenticity
    try:
        input_features = vectorizer.transform([text])
        probability = classifier.predict_proba(input_features)[0][1]  # Probability of AI-generated
    except Exception as e:
        raise RuntimeError("Failed to predict authenticity.") from e

    return float(probability)

if __name__ == "__main__":
    import argparse

    parser = argparse.ArgumentParser(description="AI-Generated Text Verifier")
    parser.add_argument("text", type=str, help="Text to analyze")
    args = parser.parse_args()

    try:
        score = verify_text(args.text)
        print(f"Probability of being AI-generated: {score:.2f}")
    except Exception as e:
        print(f"Error: {e}")