📰 News Sentiment Analysis with Text-to-Speech

📌 Overview

Analyze real-time company news, extract sentiment and key topics, and generate Hindi audio summaries using cutting-edge NLP and TTS technologies.

This project fetches the latest news articles about a given company, performs sentiment analysis, extracts key topics, and generates a Hindi spoken summary using Text-to-Speech (TTS). It includes a REST API built with Flask and a user-friendly Streamlit UI.

Features

🔍 Real-Time News Scraping from multiple online sources.

💬 Sentiment Analysis using pre-trained transformer models (Hugging Face).

📌 Key Topic Extraction with KeyBERT.

🗣️ Hindi Audio Summarization with Google Text-to-Speech (gTTS).

🌐 RESTful API for programmatic access.

🖥️ Streamlit Web App for interactive use.

🌍 Multilingual Support for broader accessibility.

🔓 Open-source and free — no paid API dependencies.

Tech Stack

Backend API: Flask (Python)

Frontend: Streamlit

Scraping: Requests, BeautifulSoup4

Sentiment Model: Transformers (BERT-like)

Topic Extraction: KeyBERT

Text-to-Speech: gTTS (Google Text-to-Speech)

Deployment Ready: Flask API & Streamlit (local/cloud)

1. Clone the Repository

git clone https://github.com/amruthadevops/NewsSummarization
cd NewsSummarization

2. Create a Virtual Environment

# For Windows
python -m venv venv
venv\Scripts\activate

# For macOS/Linux
python3 -m venv venv
source venv/bin/activate

3. Install Dependencies

pip install -r requirements.txt

Running the Application

1. Start the Flask API

python app.py

2. Start the Streamlit UI

streamlit run app.py

API Usage

Endpoint: /news_report (POST)

Request Payload:

json
{
  "company_name": "Tesla"
}

Example CURL:

curl -X POST -H "Content-Type: application/json" \ -d '{"company_name": "Tesla"}' \http://localhost:5000/news_report

Response:

{
  "summary": "Tesla stock surges amid strong quarterly results...",
  "sentiment": "Positive",
  "topics": ["electric vehicles", "earnings", "Elon Musk"]
}

🧪 Model Details

Sentiment Analysis: Transformer models fine-tuned on financial/news datasets.

Topic Extraction: KeyBERT leverages BERT embeddings for keyword generation.

Hindi TTS: gTTS supports Hindi conversion and playback of summary text.

⚠️ Assumptions & Limitations

Website structure changes may break web scraping.

Sentiment analysis may misclassify sarcasm or mixed sentiments.

TTS output is limited to gTTS’s pronunciation capabilities.

Currently optimized for Hindi language; support for others is experimental.

No rate limiting or authentication yet — not production-hardened.

🔄 Future Improvements

📰 Add more diverse and resilient news sources.

🧠 Train or fine-tune sentiment models on financial news.

🌐 Add support for other languages (e.g., Tamil, Bengali, English).

🎙️ Integrate advanced Hindi TTS like Coqui or NVIDIA FastSpeech.

☁️ Deploy on AWS/GCP with public endpoints and authentication.

Authors

• @Amrutha

• @LinkedIn

License

MIT

This project is licensed under the MIT License — feel free to use, contribute, and share!