README.md

metadata

title: Misinformation Truthteller
emoji: 📊
colorFrom: indigo
colorTo: indigo
sdk: streamlit
sdk_version: 1.41.1
app_file: app.py
pinned: false
license: mit
short_description: It is to get the model to tell truth about real world news

Misinformation Detection Tool

Overview

Misinformation has become a significant issue in today's digital age, influencing public opinion and spreading unreliable news. This project addresses the problem by building a robust Misinformation Detecting Tool using advanced Hugging Face Transformers. The system is capable of identifying whether a given news article or statement is genuine or fake.

Problem Statement

The rapid spread of misinformation through online platforms and social media has created the need for reliable tools to combat fake news. Identifying fake news manually is time-consuming and prone to bias. This project automates the detection of fake news using natural language processing (NLP) techniques, ensuring scalability and accuracy.

Objective

The objective of this project is to develop and deploy a machine learning model capable of analyzing textual data and accurately classifying it as either real or fake news. The solution is deployed using Hugging Face Transformers to make it accessible and scalable.

Features

• Deep Learning Model: Built on Hugging Face Transformers for state-of-the-art text classification.
• Scalable Deployment: Deployed on Hugging Face for seamless integration and access.
• Real-Time Prediction: Provides instant results for news articles or headlines.

Methodology

1. Data Collection:

   • Collected datasets from Kaggle and other reliable sources containing labeled news articles.

2. Data Preprocessing:

   • Cleaned and tokenized text data.
   • Removed stop words, special characters, and performed lemmatization.

3. Model Selection:

   • Used a pre-trained transformer model (e.g., BERT, RoBERTa) from Hugging Face.
   • Fine-tuned the model on the fake news dataset.

4. Training:

   • Split the dataset into training and validation sets.
   • Used PyTorch backend for training with optimization techniques.

5. Evaluation:

   • Measured performance using metrics like accuracy, precision, recall, and F1-score.
   • Validated the model with a test dataset to ensure generalizability.

6. Deployment:

   • Deployed the model on Hugging Face for public access.
   • API created for real-time predictions.

Scope

• Immediate Use: Detects fake news effectively from textual inputs such as headline or article links.
• Future Enhancements:
  • Incorporating language detection and translation for multilingual support.
  • Extending the dataset to include more diverse topics and sources.
  • Integration with video and audio analysis for multimedia content.
  • Expanded database for fact-checking and knowledge graphs.

Installation and Usage

Local Setup

1. Clone the repository:

   git lfs install
   git clone https://huggingface.co/spaces/malavika4089/misinformation-truthteller/tree/main
   cd misinformation-truthteller
   

2. Install dependencies:

   pip install -r requirements.txt
   

3. Run the script:

   streamlit run app.py
   

Access Deployed Model

The model is deployed on Hugging Face. You can access it Live link.

Dataset

The dataset used for this project was sourced from:

• Kaggle Fake And Real News Dataset

Technologies Used

• Programming Language: Python
• Libraries: Hugging Face Transformers, PyTorch, Scikit-learn, NumPy, Pandas, streamlit
• Deployment: Hugging Face Spaces,
• Tools: Colab

License

This project is licensed under the MIT License.