Repository files navigation

A web app that scrapes the latest news and summarizes them into bullet-point summaries using a fine-tuned BART model.

• Multithreading & Automated Scraping Scrapes and summarizes multiple RSS feeds in parallel, on an hourly schedule.
• Fine-Tuned ML Model: Generates detailed, concise summaries beyond the base model.
• Tooltip Feature: Select text to request definitions or explanations of terms or phrases.

NewsBrief uses multiple datasets with human-generated summaries from Hugging Face and Kaggle. These were combined into a single dataset, preprocessed, and tokenized for training. Each input is prefixed with a task instruction (summarize).

• Dataset Split: 80% training, 20% testing.
• Training uses Seq2SeqTrainingArguments with optimized learning rates and epochs. Seq2Seq Trainer handles training and evaluation. Model performance was measured withROUGE scores, comparing generated summaries with reference summaries.

The fine-tuned model is available here:Hugging Face: Article Summarizer

1. Clone the Repository

git clone https://github.com/yourusername/newsBrief.gitcd newsBrief

2. Create & activate a virtual environment

python3 -m venv venv source venv/bin/activatepip install -r requirements.txt # Install dependencies

3. Set up the database

cd backend/apipython manage.py migrate# Create a superuser for Django Admin accesspython manage.py createsuperuser

4. Configure environment variablesCreate a.env file in the project root:

GENAI_API_KEY='YOUR API KEY'

5. Run the server

python manage.py runserver

1. Install dependencies

cd frontend/newsappnpm install

2. Configure environment variablesCreate a.env file in the frontend directory:

VITE_GENAI_API_KEY='YOUR API KEY'

3. Start the development server

npm run dev