Capstone-GenAI-for-Good

Veracity Machine

The Veracity Machine is a tool built using Mesop, designed to assess the veracity of news articles. Users can upload a news article in PDF format, and then input a prompt. The system extracts the text from the article, combines it with the user prompt, and checks the veracity of the content by sending the combined input to Google's Gemini Pro AI model. The AI analyzes the content and returns an assessment or generated response based on the input.

Purpose

The Veracity Machine helps users fact-check and analyze the truthfulness of content from various news articles. It can assist in:

• Fact-checking political statements.
• Analyzing news articles for bias.
• Verifying claims made in media outlets.

Environment Creation

conda create --name <env_name> --file requirements.txt

How to run

Add your own api key for SerpApi Websearch and Google Gemini.

cd src
mesop mesop_ui.py

(May need some time for mesop to load the website)

Then click the local host address in the terminal to get to the Mesop UI.

Features

• PDF Upload: Upload news articles or other content in PDF format.
• Text Extraction: Extracts the full text from uploaded PDF files using PyPDF2.
• Prompt Combination: Combines user input (prompts) with the extracted text from the PDF for analysis.
• Google Generative AI Integration: Uses Google's Gemini Pro 1.5 model to assess the veracity of the combined input and article.
• ChromaDB Integration: Allows for storing and querying the extracted data from articles using ChromaDB.

Technologies Used

• Mesop: For creating a user-friendly interface.
• Google Generative AI (Gemini Pro 1.5): For analyzing and generating content based on the prompt and article text.
• ChromaDB: To store and retrieve article data.
• PyPDF2: For extracting text from uploaded PDFs.
• Python: Core programming language for logic and processing.

Installation

Requirements

• Python 3.7 or above
• API key for Google Generative AI.

Libraries

Install the required Python libraries using:

pip install mesop google-generativeai chromadb PyPDF2

Pipeline

1. Clone the repository:

   git clone <repository-url>
   cd <repository-directory>/src

2. Set up your environment variables: In the root directory, create a .env file and add your Google API key.

3. Add more factuality factors for Predictive AI if you want:

   cd PredictiveAI

   Change the data path or add more factors, and then:

   bash run.sh

   In order to get datasets with factors as features and average score of different sets.

4. Run the app:

   mesop mesop_ui.py

Usage

1. Upload a PDF:
   Use the "Upload PDF" button to upload a news article or report in PDF format.

2. Enter a Prompt:
   After the PDF is uploaded and its text is extracted, enter a prompt, such as a question or instruction to analyze the article.

3. Combined Input:
   The Veracity Machine will combine the extracted PDF text with your prompt for fact-checking.

4. Generate Response:
   The combined input is sent to the Gemini Pro AI, and the generated analysis or content is displayed in the output section.

5. Store and Query Articles:
   Store articles and retrieve them using ChromaDB for future reference.

src Folder Contents

• PredictiveAI/: Includes scripts and modules for predictive analysis and modeling.
• chroma_db/: Stores structured data embeddings and database files for ChromaDB management.
• saved_models/: Contains optimized pre-trained models with reduced storage requirements.
• 2024_polifact.py: Script for scraping 2024 polifacts data.
• main.py: Initial Mesop outline with foundational project functionality
• news.pdf: Sample dataset of political news in PDF format for model evaluation.
• prediction_engine.py: Core machine learning pipeline for building and evaluating predictive models.
• utils.py: Provides utility and helper functions for function calling.
• mesop_ui.py: Main script integrating PredictiveAI and GenAI models functionalities.