Josh197/ai-human-text-detection

AI Human Text Detection App

This project is a Streamlit web application that detects whether text was likely written by AI or by a human.

Problem

The app helps users evaluate pasted text or uploaded PDF/DOCX documents and returns a prediction, confidence score, explanation, and writing feedback.

Dataset

The model was trained using labeled AI-generated and human-written text samples.

Models Used

The application uses machine learning and deep learning models including SVM, Decision Tree, AdaBoost, and FNN when available.

LLM Integrations

1. 1.Prediction Explanation: explains why the model classified the text as AI-generated or human-written.
2. 2.Writing Feedback: gives suggestions based on vocabulary, sentence length, and model signals.

Results

The app displays the selected model prediction, confidence, text statistics, top TF-IDF signals, and side-by-side model comparison.

Deployment

This app is deployed on Hugging Face Spaces.